What is the impact of socio-economic factors on Criminality in France?

Did you know?

This report is available online

The public repository can also be found here: Git repository

1 Introduction

1.1 Overview and motivations

This project will analyze the factors influencing Criminality in France. We chose this subject because we are both french and criminality consistently ranks as one of the most important priority for the French population. In our study, we wanted to focus on the social and economic factors contributing to this issue. France is often regarded as one of the major world powers, and yet, it is clear that similarly to the rest of the world, there are increasing safety concerns regarding crime. This subject has also made it’s way into everyday discussions as it proves to be a polarizing topic. In the last thirty years, Politicians have often promised to tackle crime through several means, and as the population wants more to be done against it, the government has often made it one of their top priority.

In recent times, there has been a noticeable rise in criminal activities across France. Regions such as “Seine Saint Denis” in the Paris area. has seen emboldened criminals committing crimes in broad daylight, contributing to a perception of increased chaos. Similar concerns have surfaced in cities like Marseille. Additionally, it’s worth noting that the prison population has reached an all-time high in July 2023. But what are the reasons for increased criminality?

The goal of our research is to synthesize the main reasons behind criminality. Medias have often presented various reasons to explain crime, our research will look at those hints and see if they may have any statistical significance. It should be noted that data regarding “Ethnic origin” is not available, this variable is sometimes singled out by some as one of the driver of criminality and unlike in the United States, France has outlawed any form of statistics on this subject. as such, we will mostly focus on socio-economics drivers behind crime. France is facing substantial socio-economics issues, unemployment, poverty, conditions of leaving, and school level. Our project explores the link between socio-economic factors and criminality rates. Establishing such a link would give a new outlook on crime and its causes. Additionally we would like to know whether, this data could be used by the government to address public safety. In this project we will get answers regarding the core human drivers that push humans to commit crimes and how their socio-economic environment or educational path may impact their ability to commit crime.

1.2 Associated papers

https://www.institutpourlajustice.org/content/2017/12/Pauvreté-et-délinquance.pdf

https://www.bfmtv.com/politique/sondages/le-pouvoir-d-achat-reste-de-loin-la-priorite-ndeg1-des-francais-devant-la-securite-en-forte-hausse_AN-202308300557.html

https://www.leparisien.fr/politique/emmanuel-macron-donnera-ce-lundi-une-interview-sur-le-theme-de-la-securite-au-jt-de-france-3-01-10-2023-GYXJJQBWAFDEZBIX2DY3AJ3QJM.php

1.3 Research questions

The purpose of this study is to evaluate the impact of different socioeconomic factors on criminality in France. We are using for that a french government database which divide crimes in 14 categories. In our regression analysis, we decided to focus on the five most common types of crime:

1. Assault and Battery

2. Burglaries

3. Intentional damage and destructions

4. Drug use

5. Theft without violence

We will consider the following macroeconomic variables:

• Poverty rate,
• Unemployment rate
• School indicators(Success rate at an exam/Share of adults without a dimploma)
• Density of Population
• Results of previous presidential election.
• Immigration

These are the following question we will focus on:

• Can the criminality rate in France be explained using the unemployment rate?
• Can the criminality rate in France be explained using the poverty rate?
• Can the criminality rate in France be explained using education indicators ?
• Can the density of population in territories explain criminality?
• Can the immigration rate explain criminality?
• Can the previous presidential results in territories have an impact on criminality?
• Is there a correlation between these mentioned variables?

2 Data Wrangling

In this section,we will look at our raw datasets, what information they give us, how we wrangle/clean them to arrive at our final dataset. Lastly we will look at the mistakes in our data and any anomalies or outliers we may find.

Firstly, we will use a dataset with values by french state. This dataset will not be used for our Regression analysis as only having 95 observations is not enough to have a good statistical analysis.

2.1 Data by Department

2.1.1 Raw datasets

2.1.1.1 Population

There is one excel sheet per year, starting in 1975 and ending in 2023

Estimation de population au 1er janvier, par département, sexe et grande classe d'âge	...2	...3	...4	...5	...6	...7	...8	...9	...10	...11	...12	...13	...14	...15	...16	...17	...18	...19	...20
Année 2023	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA
NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA
Départements	NA	Ensemble	NA	NA	NA	NA	NA	Hommes	NA	NA	NA	NA	NA	Femmes	NA	NA	NA	NA	NA
NA	NA	0 à 19 ans	20 à 39 ans	40 à 59 ans	60 à 74 ans	75 ans et plus	Total	0 à 19 ans	20 à 39 ans	40 à 59 ans	60 à 74 ans	75 ans et plus	Total	0 à 19 ans	20 à 39 ans	40 à 59 ans	60 à 74 ans	75 ans et plus	Total
01	Ain	170727	149960	183127	109363	58760	671937	88293	75548	90943	52486	24308	331578	82434	74412	92184	56877	34452	340359
02	Aisne	125684	113454	132390	98737	52526	522791	64688	57214	65872	47242	20699	255715	60996	56240	66518	51495	31827	267076

This dataset gives us information regarding the population inside the dataset, the population is also divided by gender and age groups. We will only use the total number of people per department. It shoudl be noted that what we call our department number is a Character. This is due to the Corsica departments, split in two in 1975 and named 2A and 2B, they used to department number 20.

Source: Population by state

2.1.1.2 Crime

There is one excel sheet per department.

Index	libellé index	_2022_08	_2022_07	_2022_06	_2022_05	_2022_04	_2022_03	_2022_02	_2022_01	_2021_12	_2021_11	_2021_10	_2021_09	_2021_08	_2021_07	_2021_06	_2021_05	_2021_04	_2021_03	_2021_02	_2021_01	_2020_12	_2020_11	_2020_10	_2020_09	_2020_08	_2020_07	_2020_06	_2020_05	_2020_04	_2020_03	_2020_02	_2020_01	_2019_12	_2019_11	_2019_10	_2019_09	_2019_08	_2019_07	_2019_06	_2019_05	_2019_04	_2019_03	_2019_02	_2019_01	_2018_12	_2018_11	_2018_10	_2018_09	_2018_08	_2018_07	_2018_06	_2018_05	_2018_04	_2018_03	_2018_02	_2018_01	_2017_12	_2017_11	_2017_10	_2017_09	_2017_08	_2017_07	_2017_06	_2017_05	_2017_04	_2017_03	_2017_02	_2017_01	_2016_12	_2016_11	_2016_10	_2016_09	_2016_08	_2016_07	_2016_06	_2016_05	_2016_04	_2016_03	_2016_02	_2016_01	_2015_12	_2015_11	_2015_10	_2015_09	_2015_08	_2015_07	_2015_06	_2015_05	_2015_04	_2015_03	_2015_02	_2015_01	_2014_12	_2014_11	_2014_10	_2014_09	_2014_08	_2014_07	_2014_06	_2014_05	_2014_04	_2014_03	_2014_02	_2014_01	_2013_12	_2013_11	_2013_10	_2013_09	_2013_08	_2013_07	_2013_06	_2013_05	_2013_04	_2013_03	_2013_02	_2013_01	_2012_12	_2012_11	_2012_10	_2012_09	_2012_08	_2012_07	_2012_06	_2012_05	_2012_04	_2012_03	_2012_02	_2012_01	_2011_12	_2011_11	_2011_10	_2011_09	_2011_08	_2011_07	_2011_06	_2011_05	_2011_04	_2011_03	_2011_02	_2011_01	_2010_12	_2010_11	_2010_10	_2010_09	_2010_08	_2010_07	_2010_06	_2010_05	_2010_04	_2010_03	_2010_02	_2010_01	_2009_12	_2009_11	_2009_10	_2009_09	_2009_08	_2009_07	_2009_06	_2009_05	_2009_04	_2009_03	_2009_02	_2009_01	_2008_12	_2008_11	_2008_10	_2008_09	_2008_08	_2008_07	_2008_06	_2008_05	_2008_04	_2008_03	_2008_02	_2008_01	_2007_12	_2007_11	_2007_10	_2007_09	_2007_08	_2007_07	_2007_06	_2007_05	_2007_04	_2007_03	_2007_02	_2007_01	_2006_12	_2006_11	_2006_10	_2006_09	_2006_08	_2006_07	_2006_06	_2006_05	_2006_04	_2006_03	_2006_02	_2006_01	_2005_12	_2005_11	_2005_10	_2005_09	_2005_08	_2005_07	_2005_06	_2005_05	_2005_04	_2005_03	_2005_02	_2005_01	_2004_12	_2004_11	_2004_10	_2004_09	_2004_08	_2004_07	_2004_06	_2004_05	_2004_04	_2004_03	_2004_02	_2004_01	_2003_12	_2003_11	_2003_10	_2003_09	_2003_08	_2003_07	_2003_06	_2003_05	_2003_04	_2003_03	_2003_02	_2003_01	_2002_12	_2002_11	_2002_10	_2002_09	_2002_08	_2002_07	_2002_06	_2002_05	_2002_04	_2002_03	_2002_02	_2002_01	_2001_12	_2001_11	_2001_10	_2001_09	_2001_08	_2001_07	_2001_06	_2001_05	_2001_04	_2001_03	_2001_02	_2001_01	_2000_12	_2000_11	_2000_10	_2000_09	_2000_08	_2000_07	_2000_06	_2000_05	_2000_04	_2000_03	_2000_02	_2000_01	_1999_12	_1999_11	_1999_10	_1999_09	_1999_08	_1999_07	_1999_06	_1999_05	_1999_04	_1999_03	_1999_02	_1999_01	_1998_12	_1998_11	_1998_10	_1998_09	_1998_08	_1998_07	_1998_06	_1998_05	_1998_04	_1998_03	_1998_02	_1998_01	_1997_12	_1997_11	_1997_10	_1997_09	_1997_08	_1997_07	_1997_06	_1997_05	_1997_04	_1997_03	_1997_02	_1997_01	_1996_12	_1996_11	_1996_10	_1996_09	_1996_08	_1996_07	_1996_06	_1996_05	_1996_04	_1996_03	_1996_02	_1996_01
1	Règlements de compte entre malfaiteurs	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	1	0	0	1	1	0	0	0	0	0	0	1	1	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	1	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0
2	Homicides pour voler et à l'occasion de vols	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	2	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	2	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0
3	Homicides pour d'autres motifs	0	3	0	0	1	0	0	0	1	0	0	0	0	0	1	0	0	0	0	0	0	0	1	0	0	0	0	1	0	0	0	0	0	1	0	1	0	0	1	0	0	0	0	1	0	0	1	0	0	2	0	0	0	1	0	0	0	0	1	0	4	0	0	0	0	0	0	2	0	0	0	0	0	0	1	0	0	1	1	0	0	0	0	0	0	0	0	0	2	1	1	0	0	1	0	2	0	1	0	0	1	0	1	0	0	0	0	0	0	0	0	0	2	0	1	0	0	1	0	0	0	0	1	0	0	0	0	0	1	0	2	0	0	0	0	0	1	0	0	1	0	0	2	1	0	0	0	1	0	2	0	0	0	2	0	2	0	0	0	0	0	2	0	0	0	1	1	0	1	0	0	0	1	0	0	0	6	0	2	0	2	0	0	3	0	0	0	0	0	1	0	1	0	1	0	0	0	0	0	0	0	1	0	0	0	0	0	2	1	0	0	1	0	0	4	0	0	2	0	0	1	0	0	1	0	0	1	0	0	0	0	1	0	0	0	0	0	0	1	1	2	0	1	0	0	2	1	0	0	1	2	1	0	0	1	0	0	2	0	0	1	2	1	0	0	0	0	0	0	0	0	0	0	1	0	1	0	1	1	1	0	0	0	0	1	1	1	0	1	0	0	1	0	1	0	0	0	0	0	2	0	0	1	1	0	0	1	0	3	0	1	1	1	0	1	0	3	0	0	0
4	Tentatives d'homicides pour voler et à l'occasion de vols	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	1	0	0	0	0	0	0	0	0	0	0	0	1	0	0
5	Tentatives homicides pour d'autres motifs	0	4	0	1	1	0	1	0	2	0	0	0	0	3	1	0	0	0	1	0	0	1	1	0	3	0	1	0	3	0	0	0	2	0	0	0	0	1	0	0	0	1	0	2	0	1	0	2	0	1	0	1	3	0	1	1	0	0	0	1	1	0	2	5	2	0	1	0	1	0	1	1	0	3	1	3	4	0	0	0	0	0	1	1	0	1	0	0	2	0	2	0	0	0	0	1	0	1	0	2	0	0	0	1	0	0	2	0	0	1	0	1	0	0	0	0	0	0	2	0	0	0	0	1	0	0	1	0	0	0	2	0	0	0	0	0	2	0	0	0	0	1	2	0	1	0	1	1	0	0	3	0	1	0	1	0	1	0	0	0	0	0	1	1	7	0	0	5	0	0	0	1	0	0	0	0	0	0	1	0	0	1	0	3	1	0	1	4	0	0	0	0	0	3	0	1	1	0	0	1	1	1	0	0	1	1	0	1	1	0	0	0	0	1	0	0	0	0	0	0	2	0	0	0	2	1	0	0	1	1	2	0	0	1	0	0	0	0	0	0	0	0	2	1	0	0	0	0	0	1	4	0	1	1	0	1	0	1	1	0	0	1	0	0	0	2	2	1	0	0	0	0	0	0	0	1	0	0	0	1	0	2	1	0	3	1	1	0	1	0	0	1	1	2	2	0	1	1	0	1	0	1	0	0	1	1	0	0	2	1	0	0	0	1	0	0	1	0	1	0
6	Coups et blessures volontaires suivis de mort	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	1	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	1	0	0	0	0	0	2	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	1	0	0	0	1	0	0	0	0	0	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	1	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	1	1	1	0	0	0	0	0	3	0	1	0	1	0	0

This dataset is our main one, it tells us for example, that in the department “01” in August 2022, there was 1 case of counterfeit currency. we will be able to calculate the total number of crimes happening each month by department.

Source: Crime by state

2.1.1.3 Unemployement

Libellé	idBank	Dernière.mise.à.jour	Période	X1982.T1	X1982.T2	X1982.T3	X1982.T4	X1983.T1	X1983.T2	X1983.T3	X1983.T4	X1984.T1	X1984.T2	X1984.T3	X1984.T4	X1985.T1	X1985.T2	X1985.T3	X1985.T4	X1986.T1	X1986.T2	X1986.T3	X1986.T4	X1987.T1	X1987.T2	X1987.T3	X1987.T4	X1988.T1	X1988.T2	X1988.T3	X1988.T4	X1989.T1	X1989.T2	X1989.T3	X1989.T4	X1990.T1	X1990.T2	X1990.T3	X1990.T4	X1991.T1	X1991.T2	X1991.T3	X1991.T4	X1992.T1	X1992.T2	X1992.T3	X1992.T4	X1993.T1	X1993.T2	X1993.T3	X1993.T4	X1994.T1	X1994.T2	X1994.T3	X1994.T4	X1995.T1	X1995.T2	X1995.T3	X1995.T4	X1996.T1	X1996.T2	X1996.T3	X1996.T4	X1997.T1	X1997.T2	X1997.T3	X1997.T4	X1998.T1	X1998.T2	X1998.T3	X1998.T4	X1999.T1	X1999.T2	X1999.T3	X1999.T4	X2000.T1	X2000.T2	X2000.T3	X2000.T4	X2001.T1	X2001.T2	X2001.T3	X2001.T4	X2002.T1	X2002.T2	X2002.T3	X2002.T4	X2003.T1	X2003.T2	X2003.T3	X2003.T4	X2004.T1	X2004.T2	X2004.T3	X2004.T4	X2005.T1	X2005.T2	X2005.T3	X2005.T4	X2006.T1	X2006.T2	X2006.T3	X2006.T4	X2007.T1	X2007.T2	X2007.T3	X2007.T4	X2008.T1	X2008.T2	X2008.T3	X2008.T4	X2009.T1	X2009.T2	X2009.T3	X2009.T4	X2010.T1	X2010.T2	X2010.T3	X2010.T4	X2011.T1	X2011.T2	X2011.T3	X2011.T4	X2012.T1	X2012.T2	X2012.T3	X2012.T4	X2013.T1	X2013.T2	X2013.T3	X2013.T4	X2014.T1	X2014.T2	X2014.T3	X2014.T4	X2015.T1	X2015.T2	X2015.T3	X2015.T4	X2016.T1	X2016.T2	X2016.T3	X2016.T4	X2017.T1	X2017.T2	X2017.T3	X2017.T4	X2018.T1	X2018.T2	X2018.T3	X2018.T4	X2019.T1	X2019.T2	X2019.T3	X2019.T4	X2020.T1	X2020.T2	X2020.T3	X2020.T4	X2021.T1	X2021.T2	X2021.T3	X2021.T4	X2022.T1	X2022.T2	X2022.T3	X2022.T4	X2023.T1	X2023.T2
Taux de chômage localisé par région - France métropolitaine	1515842	29/09/2023 12:00	NA	6.7	6.8	7.0	7.0	7.0	7.0	7.2	7.5	7.9	8.3	8.5	8.7	8.8	8.8	8.8	8.8	8.7	8.8	8.8	8.9	9.0	9.0	8.8	8.8	8.7	8.6	8.5	8.3	8.1	7.9	7.8	7.8	7.7	7.7	7.6	7.6	7.6	7.7	8.0	8.2	8.4	8.6	8.8	9.0	9.2	9.5	9.8	10.1	10.3	10.4	10.3	10.1	9.8	9.6	9.5	9.6	9.9	10.1	10.2	10.2	10.3	10.3	10.3	10.1	9.9	9.8	9.7	9.8	9.8	9.8	9.5	9.1	8.7	8.3	8.0	7.7	7.4	7.3	7.3	7.4	7.5	7.5	7.5	7.6	8.0	8.1	8.0	8.4	8.6	8.4	8.5	8.5	8.3	8.5	8.7	8.7	8.8	8.6	8.6	8.0	8.2	7.8	7.7	7.1	6.9	7.0	7.1	7.4	8.2	8.9	8.9	9.1	9.0	8.9	8.9	8.8	8.8	8.7	8.9	9.0	9.1	9.4	9.4	9.8	10.0	10.1	9.9	9.8	9.8	9.8	9.9	10.1	10.0	10.2	10.0	9.9	9.9	9.7	9.6	9.8	9.3	9.2	9.2	8.7	9.0	8.8	8.6	8.4	8.5	8.1	8.1	7.9	7.7	7.0	8.8	7.9	8.0	7.7	7.7	7.2	7.1	7.1	7.1	7.0	6.9	6.9
Codes	NA		NA	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	SD
Taux de chômage localisé par région - Île-de-France	1515843	29/09/2023 12:00	NA	5.6	5.7	5.7	5.7	5.6	5.6	5.8	6.0	6.2	6.5	6.7	6.8	6.9	7.0	7.2	7.3	7.3	7.3	7.3	7.4	7.5	7.5	7.5	7.4	7.4	7.3	7.3	7.1	6.9	6.7	6.6	6.5	6.5	6.5	6.4	6.4	6.4	6.6	6.9	7.0	7.1	7.3	7.6	7.9	8.2	8.6	8.9	9.2	9.3	9.3	9.2	9.1	8.9	8.8	8.7	8.8	9.0	9.2	9.2	9.2	9.3	9.4	9.4	9.2	9.1	9.0	8.9	8.9	8.9	8.8	8.5	8.2	7.8	7.4	7.1	6.8	6.5	6.4	6.5	6.6	6.8	7.1	7.3	7.4	7.9	8.1	8.1	8.4	8.7	8.5	8.6	8.6	8.3	8.4	8.5	8.4	8.4	8.2	8.1	7.5	7.7	7.4	7.2	6.6	6.3	6.4	6.4	6.5	7.1	7.6	7.8	8.1	8.0	7.9	7.9	7.9	7.8	7.8	7.9	8.0	8.1	8.3	8.3	8.5	8.7	8.8	8.7	8.7	8.7	8.7	8.8	9.0	8.8	9.0	8.9	8.8	8.8	8.6	8.5	8.7	8.3	8.2	8.3	7.8	8.1	8.0	7.8	7.7	7.6	7.4	7.4	7.2	7.0	6.3	8.3	7.7	7.8	7.5	7.6	7.1	7.0	7.0	6.9	6.8	6.6	6.7
Codes	NA		NA	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	SD
Taux de chômage localisé par région - Centre-Val de Loire	1515847	29/09/2023 12:00	NA	5.7	5.7	5.9	6.0	6.0	6.0	6.3	6.7	7.1	7.4	7.7	8.0	8.0	8.0	8.1	8.0	8.1	8.2	8.1	8.3	8.4	8.4	8.2	8.0	7.9	7.7	7.6	7.4	7.2	7.0	6.8	6.8	6.6	6.4	6.4	6.4	6.4	6.5	6.8	7.0	7.3	7.5	7.6	7.8	8.0	8.4	8.7	9.0	9.2	9.2	9.2	8.9	8.6	8.3	8.1	8.3	8.6	8.8	9.0	9.1	9.2	9.0	8.9	8.7	8.5	8.4	8.3	8.3	8.3	8.2	7.9	7.5	7.1	6.7	6.5	6.2	6.0	5.9	5.9	6.1	6.1	6.2	6.3	6.4	6.7	7.0	7.0	7.4	7.6	7.4	7.5	7.6	7.4	7.5	7.7	7.7	7.8	7.6	7.6	7.1	7.1	6.7	6.7	6.2	5.9	6.1	6.3	6.6	7.5	8.1	8.1	8.4	8.4	8.3	8.3	8.2	8.1	8.0	8.3	8.4	8.5	8.8	9.0	9.3	9.5	9.6	9.5	9.4	9.4	9.5	9.6	9.8	9.6	9.8	9.7	9.6	9.7	9.5	9.4	9.5	9.1	9.0	8.9	8.4	8.7	8.5	8.5	8.3	8.4	8.0	8.0	7.8	7.4	6.8	8.2	7.2	7.4	7.2	7.2	6.7	6.8	6.9	6.8	6.7	6.6	6.7
Codes	NA		NA	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	SD

Source: Unemployement by state

2.1.1.4 Middle school final exam

Session	Numero.d.etablissement	Type.d.etablissement	Patronyme	Secteur.d.enseignement	Commune	Libellé.commune	Code.département	Libellé.département	Code.académie	Libellé.académie	Code.région	Libellé.région	Inscrits	Presents	Admis	Admis.sans.mention	Nombre_d_admis_Mention_AB	Admis.Mention.bien	Admis.Mention.très.bien	Taux.de.réussite
2008	0381810L	COLLEGE	GERARD PHILIPE	PUBLIC	38169	FONTAINE	038	ISERE	8	GRENOBLE	1	AUVERGNE-RHONE-ALPES	97	94	57	20	20	16	1	60,60%
2008	0381818V	LYCEE PROFESSIONNEL	-	PUBLIC	38563	VOIRON	038	ISERE	8	GRENOBLE	1	AUVERGNE-RHONE-ALPES	19	19	19	13	5	1	0	100,00%
2008	0381888W	LYCEE PROFESSIONNEL	-	PUBLIC	38509	LA TOUR-DU-PIN	038	ISERE	8	GRENOBLE	1	AUVERGNE-RHONE-ALPES	24	24	22	11	8	3	0	91,60%
2008	0382099A	LYCEE	ROGER DESCHAUX	PUBLIC	38474	SASSENAGE	038	ISERE	8	GRENOBLE	1	AUVERGNE-RHONE-ALPES	24	22	11	4	7	0	0	50,00%
2008	0382100B	COLLEGE	GRESIVAUDAN	PUBLIC	38397	SAINT-ISMIER	038	ISERE	8	GRENOBLE	1	AUVERGNE-RHONE-ALPES	193	190	174	38	57	45	34	91,50%
2008	0382104F	COLLEGE	LOUIS ARAGON	PUBLIC	38553	VILLEFONTAINE	038	ISERE	8	GRENOBLE	1	AUVERGNE-RHONE-ALPES	90	86	53	33	9	8	3	61,60%

At the end of middle school in France, sutdents must complete an exam and if they succed, the students will receive their “brevet des collèges” and can continue to high school. This would be similar to the “Certificat” in Switzerland. Here this database gives us results for each middle school in France. The first column is exam year. Column 2 to 13 tell us where and what type of school it was. After column 13, we find how many students registered, how many came and how many students obtained the “Brevet”. Lastly, we get more information regarding how well the students did. A “mention” can be obtained if the students acheive a certain average grade. Lastly, the success rate is calculated by dividing the number of students who obtained the diploma by the number of students who came the day of the exam. We will not use this sucess rate as we believe that computing the success rate by diving students who obtained the degree with registered students would give us more information regarding education levels in France.

Source: Middle school exam results by state

2.1.1.5 Election Results

Code du département	Libellé du département	Inscrits	Abstentions	% Abs/Ins	Votants	% Vot/Ins	Blancs	% Blancs/Ins	% Blancs/Vot	Nuls	% Nuls/Ins	% Nuls/Vot	Exprimés	% Exp/Ins	% Exp/Vot	Sexe...17	Nom...18	Prénom...19	Voix...20	% Voix/Ins...21	% Voix/Exp...22	Sexe...23	Nom...24	Prénom...25	Voix...26	% Voix/Ins...27	% Voix/Exp...28
1	Ain	415950	93130	22.39	322820	77.61	28852	6.94	8.94	8738	2.10	2.71	285230	68.57	88.36	M	MACRON	Emmanuel	173809	41.79	60.94	F	LE PEN	Marine	111421	26.79	39.06
2	Aisne	375791	90745	24.15	285046	75.85	22838	6.08	8.01	9067	2.41	3.18	253141	67.36	88.81	M	MACRON	Emmanuel	119202	31.72	47.09	F	LE PEN	Marine	133939	35.64	52.91
3	Allier	253479	59294	23.39	194185	76.61	18877	7.45	9.72	8522	3.36	4.39	166786	65.80	85.89	M	MACRON	Emmanuel	106579	42.05	63.90	F	LE PEN	Marine	60207	23.75	36.10
4	Alpes-de-Haute-Provence	126459	29255	23.13	97204	76.87	9671	7.65	9.95	3722	2.94	3.83	83811	66.28	86.22	M	MACRON	Emmanuel	48994	38.74	58.46	F	LE PEN	Marine	34817	27.53	41.54
5	Hautes-Alpes	109892	24895	22.65	84997	77.35	8670	7.89	10.20	2699	2.46	3.18	73628	67.00	86.62	M	MACRON	Emmanuel	47211	42.96	64.12	F	LE PEN	Marine	26417	24.04	35.88
6	Alpes-Maritimes	761780	198631	26.07	563149	73.93	47784	6.27	8.49	12414	1.63	2.20	502951	66.02	89.31	M	MACRON	Emmanuel	278407	36.55	55.35	F	LE PEN	Marine	224544	29.48	44.65

We take this sheet for the second round of election results in 2017. This second round opposed Emmanuel Macron to Marine Lepen. In this, dataset, we have the number of citizens, how many voted, blank ballots, invalid ballots and ballots for each candidate. We also have some basic information on the candidate, such as name and gender. Lastly we also have various percentages, with one being of great interest, the percentage of votes that Marine Le Pen received. We decided to take the 2017 election because with our year being 2019, seeing the “impact” of the last election on crime could yield interesting results.

Source: 2017 Election results by state

2.1.1.6 Immigration

...1	Immigrés	Non immigrés	Ensemble...4	Etrangers	Français	Ensemble...7
Ain	78968	573464	652432	592775	59657	652432
Aisne	25414	505931	531345	512545	18800	531345
Allier	17423	318552	335975	324064	11911	335975
Alpes-de-Haute-Provence	13309	150999	164308	155227	9081	164308
Hautes-Alpes	8192	133028	141220	135854	5366	141220
Alpes-Maritimes	162232	932051	1094283	972043	122240	1094283

Here we take this sheet and look at the number of immigrants by state. According to the documentation, an immigrant is a person born outside of France and living in France, while a foreigner is someone who does not have the french nationality. It should be noted that the label for foreigners and French citizens appears to have been switched in the dataframe. Fortunately, we will only use the number of immigrants for our study.

Source: Immigration numbers by departement

2.1.1.7 Population Density

Département	...2	Densité
01	Ain	113
02	Aisne	72
03	Allier	46
04	Alpes-de-Haute-Provence	24
05	Hautes-Alpes	25
06	Alpes-Maritimes	255

Source: Population Density by state

2.1.2 Data cleaning

2.1.2.1 List of departements.

We create a list of states with their numbers using the population dataset.

Code

List_depart <- List_depart[5:100, 1:2]
colnames(List_depart) <- c("Dep_number", "Dep_name")

To do this, we select our states and only the name and state code from our Population dataset. For reasons stated above, we only select metropolitan states. This means only selecting the first 96 states. We then rename our columns with easy to remember names that will be used as our key in the future.

2.1.2.2 Population by year and department

We will create 2 datasets.

Code

years_pop <- as.character(1975:2023) #every year we need to extract a sheet
Pop_by_year <- List_depart[1] #list with depart numbers 
Col_name <-  paste0("pop_", years_pop) #List to rename the columns


for (i in 1:length(years_pop)) {
  pop <- 
    read_excel(here::here("Raw_data/population par département.xls"), 
               sheet = years_pop[i]) 
  Pop_by_year <- cbind(Pop_by_year, pop[5:100, 8])#add them to the df
  colnames(Pop_by_year)[i+1] <- Col_name[i] #rename cols
}

our first dataset is the population by years. To create this, we create a vector of all the years present in the excel file. using a for loop, we open each excel sheet for each year, add them to our list of departments dataset created previously and rename it to a standard name. This dataset will be useful for our graphical representations.

we then, we take this dataset and only keep the population in 2019 and the department number. We will use this for our clustering analysis.

Code

Pop_2019 <- Pop_by_year |> 
  select(Dep_number, pop_2019)
View(Pop_2019)

2.1.2.3 Crime

Code

years_crime <- 1996:2021
Crime_total <- t(data.frame(years_crime))
colnames(Crime_total) <- Crime_total[1,]

view(Crime_total)

for (i in 1:nrow(List_depart)) {
  crimes_depart <- read_excel(here::here("Raw_data/crimes_depart.xlsx"), 
                              sheet = List_depart$Dep_number[i]) 
  #Load each datasheet with the department list
  
  crimes_depart <- crimes_depart |>
    select(-(1:10)) |> 
    summarise_all(sum) |>
    pivot_longer(cols = everything(), names_to = "Year", values_to = "Dep_number") |>
    mutate(Year = substr(Year, 2, 5)) |>
    group_by(Year) |>
    summarize(Dep_number = sum(Dep_number)) |> t()
  
  Crime_total <- rbind(Crime_total, crimes_depart[2,])
}

Crime_total <-  Crime_total[-1, ]
Crime_total <- cbind(List_depart[1], Crime_total)

We create a total number of crimes committed by department and by year.

Code

Crime_2019 <- Crime |>
  select(Dep_number, `2019`) |>
  rename(Crime_tot_2019 = `2019`)

Crime_2019 <- full_join(Crime_2019, Pop_2019, join_by("Dep_number"))
View(Crime_2019)

Crime_2019 <- Crime_2019 |>
  mutate(Crime_rate_1k = (Crime_tot_2019/pop_2019)*1000)

We then only select the year 2019 and compute the crime rate per thousand people. To do this we must add pur previously created population in 2019 dataframe.

2.1.2.4 Unemployment

This dataset is a bit trickier.

Code

dummy <- seq_len(nrow(Unemployement)) %% 2 #Dummy variable equal to 1 if the row is an odd number
Unemployement <- cbind(dummy, Unemployement)

Unemployement <- Unemployement |>
  filter(dummy == 1) |>
  select(-c(1, 3,4,5))
Unemployement <-  Unemployement[-(1:15),] # only select departments

Unemployement <- Unemployement |> 
  separate(Libellé, into = c("X", "Departement") ,sep=" - ") |>
  select(-1) 

Unemployement <- Unemployement[-(97:100), ]

First we need to remove every odd number row as they do not provide us with any information. We create a dummy variable that equals 1 if the row is an odd number, we then only select our rows with data. We then only keep the department name from the department name. We then only select metropolitan departments

Code

Unemployment_quarterly <- Unemployement |>
  pivot_longer(cols = -Dep_name, names_to = "Year", values_to = "Total") |>
  mutate(Total = as.numeric(Total)) |>
  mutate(Year = substr(Year, 2, 8)) |>
  group_by(Dep_name, Year) |>
  summarize(Total= mean(Total)) |>
  pivot_wider(names_from = Year, values_from = Total)

Unemployment_quarterly <- full_join(List_depart, Unemployment_quarterly, join_by(Dep_name))

We now need to compute the quarterly unemployment rate. To do this we use pivot_longer to be able to summarize our data by year and department. we then compute a mean unemployment rate for each year and rearrange our departments names to be inline with our list of departments using the full_join() function.

Code

Unemployment_year <- Unemployment_quarterly |>
  pivot_longer(cols = -(Dep_number:Dep_name), names_to = "Year", values_to = "Total") |>
  mutate(Year = substr(Year, 1, 4)) |>
  group_by(Dep_number, Dep_name, Year) |>
  summarize(Total = mean(Total)) |>
  pivot_wider(names_from = Year, values_from = Total)

We now compute the Unemployment by year by computing the mean of each quarter.

Code

Unemp <- Unemployment_year |> 
  select(Departement, `2019`) |>
  rename(Dep_name = Departement,
         Unemp_2019 = `2019`)

And we select only the year 2019 for our clustering and PCA analysis.

2.1.2.5 Middle school results

Code

Middle <- Middle_results |>
  filter(Session == 2019) |>
  select(Code.département, Admis, Inscrits) |>
  mutate(Pass_rate = (Admis/Inscrits)) |>
  group_by(Code.département) |>
  summarize(Pass_rate_2019 = mean(Pass_rate)) |>
  rename(Dep_number = Code.département) |>
  mutate(Dep_number = substr(Dep_number, 2, 3))

Middle <- Middle[(2:97), ]

Here we start by only taking the year 2019, we then only select useful columns, we then compute the Pass_rate which is the percentage of students who obtained the certificate. We then change the department number to be inline with our naming convention. Lastly, we select metropolitan departments.

2.1.2.6 Election results

Code

Election_by_dep <- Election_dep |>
  select(c(1,28)) |>
  rename(Dep_number = `Code du département`,
         Lepen_score = `% Voix/Exp...28`) |>
  mutate(Win_lepen = ifelse(Lepen_score>50, 1, 0))

Election_by_dep <- Election_by_dep[1:96, ]

#rename dep_number
for (i in 1:nrow(Election_by_dep)) {
  if(nchar(Election_by_dep$Dep_number[i]) == 1) {
    Election_by_dep$Dep_number[i] <- paste0("0",  Election_by_dep$Dep_number[i])
  }
}

For our Election results, we select the columns we need. Then, we select metropolitan departments. Then we need to change our department number to our naming convention to be able to use it as a key. As a reminder, our naming convention is to write single digit department with a leading zero.

2.1.2.7 Immigration

Code

Immig_2019 <- Immigration_2019 |>
  slice(1:96) |>
  select(c(1, 2, 4)) |>
  rename(Dep_name = `...1`,
         Immig_tot = Immigrés,
         pop_2019 = `Ensemble...4`) |>
  mutate(Immig_rate = Immig_tot/pop_2019) |>
  select(Dep_name, Immig_rate)

Immigration is quite simple, we select only metropolitan departments with the slice() function. We then select columns that we need, rename them in english and compute the immigration rate which is Total number of immigrants divided by the total population.

2.1.2.8 Population density

Code

Dens_2019 <- population_2019 |>
  rename(Dep_number = Département,
         Density_2019 = Densité) |>
  select(Dep_number, Density_2019) |>
  slice(1:96)

Similarly, population density is quite easy. we rename our selected columns to our naming convention, we then select only metropolitan departments using the slice() function.

2.1.2.9 Join them into a single dataset

Our last step is to join all of our previously cleaned datasets into a single dataset

Code

Full_data_dep <- full_join(List_depart, Unemp, join_by("Dep_name")) |>
  full_join(Crime_2019, join_by("Dep_number")) |>
  full_join(Middle, join_by("Dep_number")) |>
  full_join(Election_by_dep, join_by("Dep_number")) |>
  full_join(Immig_2019, join_by("Dep_name")) |>
  full_join(Dens_2019, join_by("Dep_number"))

2.2 Data by town

2.2.1 Raw datasets

2.2.1.1 Crime

CODGEO_2023	annee	classe	unité.de.compte	valeur.publiée	faits	tauxpourmille	complementinfoval	complementinfotaux	POP	millPOP	LOG	millLOG
01001	16	Autres coups et blessures volontaires	victime	ndiff	NA	NA	1,4139534883720930	0,8994774745986377	767	16	3,48000000000000e+02	16
01001	17	Autres coups et blessures volontaires	victime	ndiff	NA	NA	1,4744186046511627	0,9308908948886161	776	17	3,53730434782610e+02	17
01001	18	Autres coups et blessures volontaires	victime	ndiff	NA	NA	1,6744186046511629	1,0494250025506857	771	18	3,53730434782608e+02	18
01001	19	Autres coups et blessures volontaires	victime	diff	0	0,00000000000000e+00	NA	NA	779	19	3,59574601708852e+02	19
01001	20	Autres coups et blessures volontaires	victime	ndiff	NA	NA	1,4870689655172413	0,9798018812196120	806	20	3,66185064822065e+02	20
01001	21	Autres coups et blessures volontaires	victime	ndiff	NA	NA	1,7098214285714286	1,1468851437948424	806	20	3,66185064822065e+02	20
01001	22	Autres coups et blessures volontaires	victime	ndiff	NA	NA	1,8088888888888890	1,2238135473453471	806	20	3,66185064822065e+02	20
01004	16	Autres coups et blessures volontaires	victime	diff	39	2,76968965272353e+00	NA	NA	14081	16	7,12611602800044e+03	16
01004	17	Autres coups et blessures volontaires	victime	diff	38	2,70751692198076e+00	NA	NA	14035	17	7,21711239192120e+03	17
01004	18	Autres coups et blessures volontaires	victime	diff	46	3,23852435933540e+00	NA	NA	14204	18	7,34158489570647e+03	18

This dataset is quite big with almost 3 million rows. The first column is the CODGEO, this is a towncode by the INSEE(French institute of statistics and economics studies) the first 2 numbers are the department that the town is located in and the last 3 numbers are the town number. This first column will be very useful because it is a unique number for each town and we will use it as our key. The second column is our year, the third is the type of crime commited. The fourth column “valeur.publiée” tells us if the town publishes their crime statistics or not. It should be noted that some town are coded as publishing although the value is missing. We will dive into this later in the EDA part of our report. We then have a column telling us the total number of crimes and one for crime rate for a thousand people. The remaining columns will not be used in our analysis. The remaining LOG variables are for households in the town and the mill is a “millésime” method of census that allows a census to be conducted every year at a smaller scale, this is useful to track changes in the population year by year, but it will not be used by us.

Lastly, the rules for publishing data are the following: crime data is only published if the town records more than 5 instances of crime for 3 consecutive years. This is done to protect the privacy of the people involved and because the statistical analysis of such small data would not yield satisfying results.

Source: Crime by town

2.2.1.2 Density

X1	X2	X3	X4	X5	X6	X7	X8	X9	X10	X11	X12	X13	X14	X15	X16	X17	X18	X19	X20	X21	X22	X23	X24	X25	X26	X27
1	01	ozan	OZAN	ozan	Ozan	O250	OSN	01190	284	01284	2	26	6	618	469	500	93	6.60	4.917	46.38	2866	51546	+45456	462330	170	205
2	01	cormoranche-sur-saone	CORMORANCHE-SUR-SAONE	cormoranche sur saone	Cormoranche-sur-Saône	C65652625	KRMRNXSRSN	01290	123	01123	2	27	6	1058	903	1000	107	9.85	4.833	46.23	2772	51379	+44953	461427	168	211
3	01	plagne-01	PLAGNE	plagne	Plagne	P425	PLKN	01130	298	01298	4	03	6	129	83	100	20	6.20	5.733	46.18	3769	51324	+54342	461131	560	922
4	01	tossiat	TOSSIAT	tossiat	Tossiat	T230	TST	01250	422	01422	2	25	6	1406	1111	1400	138	10.17	5.317	46.13	3309	51268	+51854	460828	244	501
5	01	pouillat	POUILLAT	pouillat	Pouillat	P430	PLT	01250	309	01309	2	33	6	88	58	100	14	6.23	5.433	46.33	3435	51475	+52542	461938	333	770
6	01	torcieu	TORCIEU	torcieu	Torcieu	T620	TRS	01230	421	01421	1	28	6	698	643	700	65	10.72	5.400	45.92	3398	51025	+52343	455521	257	782

This dataframe contains various information regarding french towns. Here, can be found the size of each town, the population but also the altitude for example. Here we will use the population in 2010 and the size to compute the density.

Source: Density by town

2.2.1.3 Population by town

NIVGEO	CODGEO	LIBGEO	SEXE	AGEPYR10	NB
COM	01001	L'Abergement-Clémenciat	1	00	14.813
COM	01001	L'Abergement-Clémenciat	2	00	13.905
COM	01001	L'Abergement-Clémenciat	1	03	10.848
COM	01001	L'Abergement-Clémenciat	2	03	6.856
COM	01001	L'Abergement-Clémenciat	1	06	23.656
COM	01001	L'Abergement-Clémenciat	2	06	32.462

This dataset is quite simple, we have a column indicating the level, here is is towns, the CODGEO mentioned above, the town name, followed by the gender and age group. finally we have the number of inhabitants that fit these criterias.

Source: Density by town

2.2.1.4 Population by diploma

Région en géographie courante	Département en géographie courante	Commune en géographie courante	Indicateur de stabilité de la commune	Département en géographie 2021	Libellé de commune	Aucun diplôme Hommes 16 à 24 ans RP2019	Aucun diplôme Hommes 25 ans ou plus RP2019	Aucun diplôme Femmes 16 à 24 ans RP2019	Aucun diplôme Femmes 25 ans ou plus RP2019	Diplôme de niveau CEP Hommes 16 à 24 ans RP2019	Diplôme de niveau CEP Hommes 25 ans ou plus RP2019	Diplôme de niveau CEP Femmes 16 à 24 ans RP2019	Diplôme de niveau CEP Femmes 25 ans ou plus RP2019	Diplôme de niveau BEPC Hommes 16 à 24 ans RP2019	Diplôme de niveau BEPC Hommes 25 ans ou plus RP2019	Diplôme de niveau BEPC Femmes 16 à 24 ans RP2019	Diplôme de niveau BEPC Femmes 25 ans ou plus RP2019	Diplôme de niveau CAP-BEP Hommes 16 à 24 ans RP2019	Diplôme de niveau CAP-BEP Hommes 25 ans ou plus RP2019	Diplôme de niveau CAP-BEP Femmes 16 à 24 ans RP2019	Diplôme de niveau CAP-BEP Femmes 25 ans ou plus RP2019	Diplôme de niveau bac général ou technique Hommes 16 à 24 ans RP2019	Diplôme de niveau bac général ou technique Hommes 25 ans ou plus RP2019	Diplôme de niveau bac général ou technique Femmes 16 à 24 ans RP2019	Diplôme de niveau bac général ou technique Femmes 25 ans ou plus RP2019	Diplôme universitaire de 1er cycle Hommes 16 à 24 ans RP2019	Diplôme universitaire de 1er cycle Hommes 25 ans ou plus RP2019	Diplôme universitaire de 1er cycle Femmes 16 à 24 ans RP2019	Diplôme universitaire de 1er cycle Femmes 25 ans ou plus RP2019	Diplôme universitaire de 2ème ou 3ème cycle Hommes 16 à 24 ans RP2019	Diplôme universitaire de 2ème ou 3ème cycle Hommes 25 ans ou plus RP2019	Diplôme universitaire de 2ème ou 3ème cycle Femmes 16 à 24 ans RP2019	Diplôme universitaire de 2ème ou 3ème cycle Femmes 25 ans ou plus RP2019
RR	DR	CR	STABLE	DR21	LIBELLE	dpx_rec0s1age2_rec1rpop2019	dpx_rec0s1age2_rec2rpop2019	dpx_rec0s2age2_rec1rpop2019	dpx_rec0s2age2_rec2rpop2019	dpx_rec1s1age2_rec1rpop2019	dpx_rec1s1age2_rec2rpop2019	dpx_rec1s2age2_rec1rpop2019	dpx_rec1s2age2_rec2rpop2019	dpx_rec2s1age2_rec1rpop2019	dpx_rec2s1age2_rec2rpop2019	dpx_rec2s2age2_rec1rpop2019	dpx_rec2s2age2_rec2rpop2019	dpx_rec3s1age2_rec1rpop2019	dpx_rec3s1age2_rec2rpop2019	dpx_rec3s2age2_rec1rpop2019	dpx_rec3s2age2_rec2rpop2019	dpx_rec4s1age2_rec1rpop2019	dpx_rec4s1age2_rec2rpop2019	dpx_rec4s2age2_rec1rpop2019	dpx_rec4s2age2_rec2rpop2019	dpx_rec5s1age2_rec1rpop2019	dpx_rec5s1age2_rec2rpop2019	dpx_rec5s2age2_rec1rpop2019	dpx_rec5s2age2_rec2rpop2019	dpx_rec6s1age2_rec1rpop2019	dpx_rec6s1age2_rec2rpop2019	dpx_rec6s2age2_rec1rpop2019	dpx_rec6s2age2_rec2rpop2019
84	01	001	1	01	Abergement-Clémenciat	0	31.273727679978059	0	26.319004906244121	0	36.504516945208763	0	42.168740366590413	4.8204307590319528	25.195300294679452	0	25.897214253562552	0	72.042804073821756	10.047478793216261	66.528046044671314	9.9094455474678735	51.11132775508888	4.9547227737339368	50.935251110259628	10.185512038964646	25.624888993112183	0	35.403682355195734	0	45.715893991511749	0	30.43378047534657
84	01	002	1	01	Abergement-De-Varey	0	0	0	5.5320444734237153	0	5.1374280962262988	0	10.402372543219041	0	5.2011862716095205	0	0	0	31.540359468699528	0	10.466921990976699	0	31.273259442268557	0	26.136622618416695	0	15.73425007519765	0	10.596821978971352	0	10.532272531213694	0	21.129094510185048
84	01	003	0	01	Amareins	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA
84	01	004	1	01	Ambérieu-En-Bugey	76.83122693441176	643.20453199840733	35.464379173014933	860.73512425945796	0	209.96947893832353	0	432.8132259662828	42.914918277527164	207.23820326380567	52.610753884473795	355.55757783628127	181.41495713237916	1382.3332572324048	122.86736219321433	1245.913072797086	156.88926046556423	822.49605569888286	188.34449242818874	832.16467035277583	50.77482939272798	638.97964337399492	50.046583285767468	536.94746995634785	39.280654212696781	593.41118637421562	47.035995912555691	837.01126571351983
84	01	005	1	01	Ambérieux-En-Dombes	0	83.521324354657963	0	44.217171717171972	0	19.652076318743237	0	63.869248035915504	9.8260381593715458	39.304152637485977	0	49.130190796858017	4.913019079685748	240.73793490460261	4.913019079685748	167.04264870931641	4.9130190796857969	88.434343434343873	9.8260381593715458	152.30359147025854	0	54.043209876543123	4.9130190796856503	93.347362514029356	0	93.347362514029072	4.913019079685748	83.521324354657537

In this dataset, column 1 to 6 are the location of the town, with indicators such as the region, the departement and the town number. Then the population is divided by gender and by diploma obtained, we will only focus on the total population over 25 who doesn’t have any diploma. The whole excel files gives us this information for every year starting in 1968. We only take the year 2019.

Source: Population by diploma by town

2.2.1.5 2017 Election results

Code du département	Libellé du département	Code de la commune	Libellé de la commune	Inscrits	Abstentions	% Abs/Ins	Votants	% Vot/Ins	Blancs	% Blancs/Ins	% Blancs/Vot	Nuls	% Nuls/Ins	% Nuls/Vot	Exprimés	% Exp/Ins	% Exp/Vot	N°Panneau...19	Sexe...20	Nom...21	Prénom...22	Voix...23	% Voix/Ins...24	% Voix/Exp...25	N°Panneau...26	Sexe...27	Nom...28	Prénom...29	Voix...30	% Voix/Ins...31	% Voix/Exp...32
1	Ain	1	L'Abergement-Clémenciat	598	100	16.72	498	83.28	37	6.19	7.43	8	1.34	1.61	453	75.75	90.96	1	M	MACRON	Emmanuel	272	45.48	60.04	2	F	LE PEN	Marine	181	30.27	39.96
1	Ain	2	L'Abergement-de-Varey	209	32	15.31	177	84.69	21	10.05	11.86	6	2.87	3.39	150	71.77	84.75	1	M	MACRON	Emmanuel	93	44.50	62.00	2	F	LE PEN	Marine	57	27.27	38.00
1	Ain	4	Ambérieu-en-Bugey	8586	2312	26.93	6274	73.07	601	7.00	9.58	217	2.53	3.46	5456	63.55	86.96	1	M	MACRON	Emmanuel	3213	37.42	58.89	2	F	LE PEN	Marine	2243	26.12	41.11
1	Ain	5	Ambérieux-en-Dombes	1172	259	22.10	913	77.90	77	6.57	8.43	24	2.05	2.63	812	69.28	88.94	1	M	MACRON	Emmanuel	423	36.09	52.09	2	F	LE PEN	Marine	389	33.19	47.91
1	Ain	6	Ambléon	99	23	23.23	76	76.77	17	17.17	22.37	0	0.00	0.00	59	59.60	77.63	1	M	MACRON	Emmanuel	30	30.30	50.85	2	F	LE PEN	Marine	29	29.29	49.15
1	Ain	7	Ambronay	1880	336	17.87	1544	82.13	138	7.34	8.94	31	1.65	2.01	1375	73.14	89.05	1	M	MACRON	Emmanuel	774	41.17	56.29	2	F	LE PEN	Marine	601	31.97	43.71

This dataset is very similar to the one for departments.

Source: 2017 Election results by town

2.2.1.6 Poverty by town

CODGEO	TP4019	TP5019	TP6019	TP60Q219	TP60IP19	AGE1TP4019	AGE1TP5019	AGE1TP6019	AGE1TP60Q219	AGE1TP60IP19	AGE2TP4019	AGE2TP5019	AGE2TP6019	AGE2TP60Q219	AGE2TP60IP19	AGE3TP4019	AGE3TP5019	AGE3TP6019	AGE3TP60Q219	AGE3TP60IP19	AGE4TP4019	AGE4TP5019	AGE4TP6019	AGE4TP60Q219	AGE4TP60IP19	AGE5TP4019	AGE5TP5019	AGE5TP6019	AGE5TP60Q219	AGE5TP60IP19	AGE6TP4019	AGE6TP5019	AGE6TP6019	AGE6TP60Q219	AGE6TP60IP19	TME1TP4019	TME1TP5019	TME1TP6019	TME1TP60Q219	TME1TP60IP19	TME2TP4019	TME2TP5019	TME2TP6019	TME2TP60Q219	TME2TP60IP19	TME3TP4019	TME3TP5019	TME3TP6019	TME3TP60Q219	TME3TP60IP19	TME4TP4019	TME4TP5019	TME4TP6019	TME4TP60Q219	TME4TP60IP19	TME5TP4019	TME5TP5019	TME5TP6019	TME5TP60Q219	TME5TP60IP19	TOL1TP4019	TOL1TP5019	TOL1TP6019	TOL1TP60Q219	TOL1TP60IP19	TOL2TP4019	TOL2TP5019	TOL2TP6019	TOL2TP60Q219	TOL2TP60IP19	TLD2TP4019	TLD2TP5019	TLD2TP6019	TLD2TP60Q219	TLD2TP60IP19	TLD3TP4019	TLD3TP5019	TLD3TP6019	TLD3TP60Q219	TLD3TP60IP19	TYM1TP4019	TYM1TP5019	TYM1TP6019	TYM1TP60Q219	TYM1TP60IP19	TYM2TP4019	TYM2TP5019	TYM2TP6019	TYM2TP60Q219	TYM2TP60IP19	TYM3TP4019	TYM3TP5019	TYM3TP6019	TYM3TP60Q219	TYM3TP60IP19	TYM4TP4019	TYM4TP5019	TYM4TP6019	TYM4TP60Q219	TYM4TP60IP19	TYM5TP4019	TYM5TP5019	TYM5TP6019	TYM5TP60Q219	TYM5TP60IP19	TYM6TP4019	TYM6TP5019	TYM6TP6019	TYM6TP60Q219	TYM6TP60IP19	OPR1TP4019	OPR1TP5019	OPR1TP6019	OPR1TP60Q219	OPR1TP60IP19	OPR2TP4019	OPR2TP5019	OPR2TP6019	OPR2TP60Q219	OPR2TP60IP19	OPR3TP4019	OPR3TP5019	OPR3TP6019	OPR3TP60Q219	OPR3TP60IP19	OPR4TP4019	OPR4TP5019	OPR4TP6019	OPR4TP60Q219	OPR4TP60IP19	OPR5TP4019	OPR5TP5019	OPR5TP6019	OPR5TP60Q219	OPR5TP60IP19
01004	16	22	28	7980	0.385	17	23	30	7950	0.388	23	31	37	7500	0.423	19	28	35	8060	0.379	14	19	23	7180	0.447	NA	10	15	9330	0.281	NA	NA	NA	NA	NA	12	17	22	7750	0.404	9	12	16	7930	0.389	13	20	26	8430	0.351	14	21	27	8420	0.352	34	45	53	7560	0.418	5	6	9	9460	0.271	26	35	43	7520	0.421	37	49	58	7170	0.448	14	19	28	8580	0.34	NA	NA	20	7030	0.459	NA	18	24	8430	0.351	NA	NA	9	9160	0.295	17	24	31	8280	0.362	31	40	49	6920	0.467	NA	NA	NA	NA	NA	12	18	25	8960	0.310	67	NA	NA	5980	0.539	NA	NA	NA	NA	NA	8	12	17	8760	0.326	NA	NA	NA	NA	NA
01007	NA	9	13	9400	0.276	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	38	8990	0.308	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	12	10020	0.229	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA
01014	8	13	19	9390	0.277	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	25	9690	0.254	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	34	9430	0.274	NA	NA	10	9790	0.246	17	27	37	8920	0.313	NA	30	41	9050	0.303	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	13	20	9480	0.270	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	10	17	10010	0.229	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA
01024	NA	8	12	9430	0.274	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	8	10000	0.230	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	11	8880	0.316	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	11	10410	0.198	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA
01025	5	7	11	9020	0.305	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	7	9760	0.249	NA	NA	34	8310	0.360	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	10	8400	0.353	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	9	9690	0.254	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA
01031	17	23	29	7930	0.389	NA	NA	NA	NA	NA	NA	37	44	7660	0.410	23	32	40	8020	0.383	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	35	47	56	7310	0.437	NA	NA	NA	NA	NA	28	37	47	7850	0.396	30	41	52	7920	0.390	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	19	26	34	8290	0.361	NA	NA	NA	6640	0.489	NA	NA	NA	NA	NA	13	20	28	8780	0.324	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA

This dataset does not have easily understable column names. The first is CODGEO, the town code. The columns starting with “TP” indicate poverty rates. the number afterwards indicate the rate at which people are considered poor. For example, with 60% which is the commonly used cutoff for poverty, the number means that a certain percentage of people earn only 60% of the median revenue in France. The column “TP60IP19” is the poverty intensity. This is the difference between the poverty threshold and the median revenue of people considered poor. The remaining columns divide those values by age.

Source: Poverty by town

2.2.1.7 Immigration by town

NIVGEO	CODGEO	LIBGEO	AGE4	IMMI	SEXE	NB
COM	01001	L'Abergement-Clémenciat	00	2	1	82.648
COM	01001	L'Abergement-Clémenciat	15	1	1	1.019
COM	01001	L'Abergement-Clémenciat	15	2	1	39.016
COM	01001	L'Abergement-Clémenciat	25	1	1	3.965
COM	01001	L'Abergement-Clémenciat	25	2	1	145.677
COM	01001	L'Abergement-Clémenciat	55	1	1	7.300

Here we have the number of immigrants by town, age and gender. The first 3 columns are town indicators, code and name. the AGE4 variable indicate the age group. IMMI is the immigration status, 1 for immigrant, and 2 for non immigrants. SEXE is the gender, 1 for men and 2 for women and NB is the total number.

2.2.2 Cleaning

2.2.2.1 Crime

Code

Type_per_year <- Crimes |> 
  filter(valeur.publiée == "diff") |>
  filter(!grepl("^97", CODGEO_2023)) |>
  select(CODGEO_2023, annee, classe, faits, tauxpourmille)

Type_per_year <- cbind(Type_per_year[1], Type_per_year)
colnames(Type_per_year) <- c("Departement", "Town", "Year", "Type", "Number", "Rate_per_1k")

Type_per_year <- Type_per_year |> 
  mutate(Departement = substr(Departement, 1, 2)) |>
  mutate(Rate_per_1k = str_replace_all(Rate_per_1k, ",", ".")) |>
  mutate(Rate_per_1k = as.numeric(Rate_per_1k)) |>
  group_by(Departement, Town, Year, Number, Type) |>
  summarize(Rate_per_1k = mean(Rate_per_1k, na.rm = TRUE), ) |> #omit the missing values because some are missing even though they were indicated as published
  group_by(Departement, Town, Year, Type, Rate_per_1k) |>
  summarize(Number = sum(Number))

We use this code to get the data by type and by year for each town. First, we only select values indicated as publishing their data (“diff”). We then only take metropolitan departments, by filtering every row except those starting with 97(the indicator for outer seas department). We then duplicate the town code to create a Department column. we rename the columns for more clarity. we then create our department column by substr() the first 2 digits of the town code. We now need to compute the rate for a thousand. To do this, we first must replace every comma by a point, that is because European notation of numbers uses a comma instead of a point for decimal numbers. we then use group_by() and summarize() to get the mean rate. and use the same functions to get the total number of crimes. This dataframe will be useful for our plots

Code

Crime_2019_town <- Type_per_year |>
  filter(Year == 19) |>
  group_by(Departement, Town) |>
  summarize(Rate_per_1k = mean(Rate_per_1k),
         Total_crime = sum(Number)) |>
  rename(Town_code = Town, Dep_number = Departement)

We then summarize the rate value by town to use in our regression

Code

#Create a dataset where we look at the rate for each type of crime
Crime_type_town <- Type_per_year |>
  filter(Year == 19) |>
  rename(Town_code = Town, Dep_number = Departement) |>
  select(Dep_number, Town_code, Type, Rate_per_1k) |>
  pivot_wider(names_from = Type, values_from = Rate_per_1k) |>
  select(Dep_number, Town_code, "Cambriolages de logement", 
         "Coups et blessures volontaires", 
         "Destructions et dégradations volontaires", 
         "Usage de stupéfiants",
         "Vols sans violence contre des personnes")

Crime_type_town[is.na(Crime_type_town)] <- 0

We also create a dataset where we keep the type of crime commited. To do this, we simply do a pivot_wider() and select our types of crime. In this instance we only focus on the 5 most commited crimes. Since some cities do not have every type of crime present, we are left with a substantial amount of NA values. We decide to replace our NA’s with zeroes since an NA value indicates that there was less than 5 instances of a crime.

2.2.2.2 Density

Code

Density_town <- Dense |>
  select(c(4, 11, 15, 19)) |>
  rename(Town_name = `X4`, 
         Town_code = `X11`,
         Pop_2012 = `X15`,
         Size = `X19`) |>
  filter(!str_detect(Town_code, "^97")) |>
  mutate(Density_2019 = Pop_2012/Size) |>
  select(Town_code, Town_name, Density_2019)

To do this, we first select the columns we need. In that case, the name of the town, the Town_code, the population and the size. We then rename them to our standard, only select metropolitan departments and compute the density(Population/Size) lastly, we select only our important columns.

2.2.2.3 Population

Code

Pop_by_town <- Pop_2019 |>
  group_by(CODGEO) |>
  summarise(Total = sum(NB)) |>
  rename(Town_code = CODGEO, Total_pop = Total)
view(Pop_by_town)

The cleaning for this dataset is very simple, we compute the sum of every age group and gender group to find the total number of inhabitants by town.

2.2.2.4 Population without a diploma

Code

Pop_no_diploma <- Pop_16_no_diploma |>
  select(c(2, 3, 6, 8, 10)) |>
  slice(-1) |>
  rename(Dep_number = `Département\nen géographie courante`,
         Town_number = `Commune\nen géographie courante`,
         Town_name = `Libellé de commune`,
         no_diploma_M = `Aucun diplôme\nHommes\n25 ans ou plus\nRP2019`,
         no_diploma_F = `Aucun diplôme\nFemmes\n25 ans ou plus\nRP2019`) |>
  group_by(Dep_number, Town_number) |>
  summarize(no_diploma = sum(as.numeric(no_diploma_F), as.numeric(no_diploma_M), na.rm = TRUE)) |>
  mutate(Town_code = paste0(Dep_number, Town_number)) |>
  select(Dep_number, Town_code, no_diploma) |>
  filter(!str_detect(Dep_number, "^97")) |>
  left_join(Pop_by_town, join_by(Town_code)) |>
  na.omit() |>
  mutate(No_diploma_rate1k = (no_diploma/Total_pop)*1000)

Paris_tot <- Pop_no_diploma |>
  filter(Dep_number == "75") |>
  summarize(No_diploma_rate1k = mean(No_diploma_rate1k)) |>
  rename(Town_code = Dep_number) |>
  mutate(Town_code = "75056")

Pop_no_diploma <- Pop_no_diploma |>
  group_by(Town_code, No_diploma_rate1k) |>
  select(Town_code, No_diploma_rate1k)

Pop_no_diploma <- rbind(Pop_no_diploma, Paris_tot)

To have our final data, we first select our needed columns. we then remove the first row and rename the columns. we then compute the total number of people without a diploma and make sur to remove any NA values. we then select only metropolitan departments, with filter() and str_detect. We then compute a left join with the dataframe Pop_by_town to get the population by town. afterwards, we compute the rate at which people do not have a diploma for a thousand people by dividing the total number of people without any diploma by the population of the town and multiply it by a thousand.

Lastly, we have noticed that this dataset provides us with data from every subdivision in the city of Paris, while this level of precision could be welcome, all of our other datasets counts Paris as a single city. As such, we must aggregate our rate for the city of Paris. To do this, we create a new df called Paris_tot, we then summarize all of it. We add the row at the end of our previous dataframe. It is unordered, but will become ordered when we use the join function later on.

2.2.2.5 2017 election results.

Code

Vote_2017 <- Presidentielle_2017_Resultats_Communes_Tour_2_c |>
  select(c(1, 2, 3, 25, 32)) #select ony the values important to us in that case, the department, the towncode and the result
Vote_2017 <- Vote_2017[1:35281, ]
colnames(Vote_2017) <- c("Department", "Dep_name", "Town_num", "Macron", "Lepen")

#Need to add the Corse departement number(2A/2B)
Vote_2017$Department <- as.character(Vote_2017$Department)

for (i in 1:nrow(Vote_2017)) {
  if (Vote_2017$Dep_name[i] == "Corse-du-Sud") {
    Vote_2017$Department[i] <-  "2A"
  }
  if (Vote_2017$Dep_name[i] == "Haute-Corse") {
    Vote_2017$Department[i] <-  "2B"
  }  
}

#Now let's standardize the notation of departments and town codes by adding zeroes before.
for (i in 1:nrow(Vote_2017)) {
  if (nchar(Vote_2017$Department[i]) == 1) {
    Vote_2017$Department[i] <- paste0("0", Vote_2017$Department[i])
  }
  if (nchar(Vote_2017$Town_num[i])<=2) {
    if (nchar(Vote_2017$Town_num[i]) <= 1) { 
           Vote_2017$Town_num[i] <- paste0("00", Vote_2017$Town_num[i])
    }
           else {Vote_2017$Town_num[i] <- paste0("0", Vote_2017$Town_num[i])
           }
  }
}

#Create the town key
Vote_2017 <- Vote_2017 |> 
  mutate(Town_code = paste0(Department, Town_num)) |>
  mutate(Win_Lepen = ifelse(Lepen>50, 1, 0)) |>
  select(Department, Town_code, Macron, Lepen, Win_Lepen)#Create a dummy variable that equals 1 if LePen "won" the town (scored more than 50%)

we first select our needed columns. then, we only select metropolitan departments and rename our columns. the next step is to code the Corsica departments numbers. Corsica is split into 2 departments. 2A: “Corse-du-Sud” and 2B: “Haute-Corse”. these 2 departments are found between the department 19 and department 21. They used to be one single department butt were split in 2 in 1975. In our database, only the name can be found. we need the number. we use a for loop to code it into the Departement column. We then standardize our Departments numbers and town codes by adding leading zeroes. This then allows us to create our town_number key. Finally, we code a dummy variable that equals 1 if Marine Lepen had more than 50% of the votes in a town.

2.2.2.6 Poverty

Code

Povr_2019 <- Pov_2019 |>
  select(c(1, 4, 6)) |>
  na.omit() |>
  rename(Town_code = CODGEO,
         Povrety_2019 = TP6019,
         Intensity_povrety = TP60IP19)

Poverty is quite simple we select our columns, rename them and we do not forget to omit values that were blank

2.2.2.7 Immigration

Code

Immig_2019_town <- Immig_by_town |>
  select(CODGEO, IMMI, NB) |>
  rename(Town_code = CODGEO) |>
  filter(IMMI == 1) |>
  filter(!str_detect(Town_code, "^97")) |>
  group_by(Town_code) |>
  summarize(NB = sum(NB)) |>
  left_join(Pop_by_town, join_by(Town_code)) |>
  mutate(Immig_rate = (NB/Total_pop)*1000 ) |>
  na.omit() |>
  select(Town_code, Immig_rate)

Here we apply the same framework we have applied before. We select our needed columns, rename them. we then filter to only have the total of immigrants. then, we add the population from our previously created population dataset using left_join() we then compute the immigration rate, omit missing values.

2.2.2.8 Join them into a single dataframe

Finally, we join all of those datasets into a single one. To do this, we first use full_join() to see how values are missing.

Code

Everything_by_town <- Crime_2019_town |>
  full_join(Pop_by_town, join_by(Town_code)) |> 
  full_join(Density_town, join_by(Town_code)) |>
  full_join(Vote_2017, join_by(Town_code)) |>
  full_join(Povr_2019, join_by(Town_code)) |>
  full_join(Pop_no_diploma, join_by(Town_code)) |>
  full_join(Immig_2019_town, join_by(Town_code)) |>
  left_join(Unemp, join_by(Dep_number))

summary(Everything_by_town)

As we can see, Poverty is a smaller dataset. This may be a problem, but upon further inspection, we see that the smallest population is 1’700 inhabitants. And we still have around 5’000 observations. We previously thought about removing every observation where the number of inhabitants of a town was smaller than 200, because of the poor statistical analysis that would stem from such extreme values. the INSEE institute were most of our data has been collected recommends not doing any analysis on towns of less than 200 inhabitants. As such, the reduced number of observations with the poverty dataset is not a problem. We also add our unemployment numbers from our by_dep dataset because we were unable to find accurate unemployment numbers by department.

Code

Everything_by_town_clean <- Everything_by_town |>
  na.omit() |>
  select(Dep_number, 
         Town_code, 
         Town_name, 
         Total_pop, Rate_per_1k, Density_2019, Lepen, Win_Lepen, Povrety_2019, Intensity_povrety, No_diploma_rate1k, Immig_rate, Unemp_2019)

then, we omit our missing values, select only our very important values and we have our final dataset. that we may use for EDA and most importantly for our Regression analysis. Were we will predict the crime rate of a town using our independent variables.

We then create a dataset for each type of crime per town.

Code

Crime_per_type_town <- Crime_type_town |>
  full_join(Everything_by_town_clean, join_by(Town_code)) |>
  select(Dep_number.x, Town_code, Town_name, 
         Total_pop, Total_crime, Rate_per_1k, 
         "Cambriolages de logement", "Coups et blessures volontaires", "Destructions et dégradations volontaires", 
         "Usage de stupéfiants", "Vols sans violence contre des personnes",
         Density_2019, Lepen, Win_Lepen, 
         Povrety_2019, Intensity_povrety, No_diploma_rate1k, 
         Immig_rate, Unemp_2019) |>
  na.omit() |>
  rename(Assault = `Coups et blessures volontaires`,
         Burglary = `Cambriolages de logement`,
         Damage = `Destructions et dégradations volontaires`,
         Drugs = `Usage de stupéfiants`,
         Theft = `Vols sans violence contre des personnes`,
         Dep_number = Dep_number.x
         ) |>
  select(-Year) |>
  mutate(Lepen= Lepen/100,
         Povrety_2019 = Povrety_2019/100,
         No_diploma_rate1k = No_diploma_rate1k/1000,
         Immig_rate = Immig_rate/1000,
         Unemp_2019 = Unemp_2019/100)

2.3 Final results

We finally have these datasets.

The Department level:

The town level:

Dep_number	Town_code	Rate_per_1k	Total_crime	Total_pop	Town_name	Density_2019	Department	Macron	Lepen	Win_Lepen	Povrety_2019	Intensity_povrety	No_diploma_rate1k	Immig_rate	Dep_name	Unemp_2019
01	01001	0.000	0	779	L'ABERGEMENT-CLEMENCIAT	49.15	01	60.04	39.96	0	NA	NA	73.93	26.30	Ain	6.075
01	01002	0.000	0	256	L'ABERGEMENT-DE-VAREY	24.15	01	62.00	38.00	0	NA	NA	21.61	12.34	Ain	6.075
01	01004	4.659	722	14134	AMBERIEU-EN-BUGEY	562.40	01	58.89	41.11	0	28	0.385	106.41	111.73	Ain	6.075
01	01005	3.736	12	1751	AMBERIEUX-EN-DOMBES	101.51	01	52.09	47.91	0	NA	NA	72.95	40.97	Ain	6.075
01	01006	0.000	0	112	AMBLEON	19.73	01	50.85	49.15	0	NA	NA	45.45	72.73	Ain	6.075
01	01007	5.021	74	2800	AMBRONAY	70.40	01	56.29	43.71	0	13	0.276	42.86	36.07	Ain	6.075

As we can see, this dataset includes many towns of very few inhabitants. And many Missing values have been introduced into this dataset. As such, we will omit those missing values that were mostly located in the Poverty department. We are left with this final clean dataset

The smallest town in this dataset is 1711. This will be a much better dataset to run our regressions on.

3 EDA

3.1 Visualization

3.1.1 Crime

3.1.1.1 Evolution of the total crime rate

In this graphic we chose to compute the crime rate per hundred persons in the 5 departments with the biggest and the 5 with the lowest crime rate.

In this graphic we chose to compute the crime rate per hundred persons in the 5 departments with the highest crime rate and those with the lowest crime rate. The first thing we can observe is the major disparity between the average crime number in those departments. In fact the highest crime rate in the Paris is almost 5 times higher than the Aveyron 30.85 vs 145.25 It’s important to note that the departments with the higher crime rate are the one encompassing the biggest cities in France: Paris and it’s suburbs of “Seine-Saint-Denis” and “Hauts-de-Seine”, Marseille is in the “Bouches-du_Rhône”, Lyon is located in the “Rhône”. We can start to understand that density may be a key factor in determining crime rate.

Evolution of Crime in the 5 most criminal departments

Here we can observe that the total crimes in the most criminal departments are lower than what they were twenty years ago in the city of Paris (75) and Marseille (13) but are sensibly the same or higher in the peripheral areas of Paris (92/93) and in the city of Lyon (69). This can possibly be explained by the huge increase of density of population in the peripheral areas of big cities in France.

This graph represents the annual amount of crimes in the 5 least criminal departments. Here we can assume that the number of crimes in the safest departments is quite stable through the years. We also observe the small variation due to COVID-19 Pandemic

3.1.1.2 Types of crime commited

We start by looking at the evolution of the most common types of crime by year in France.

Here we can see that the most common crime in France is deliberate destruction and damage. The effect of the COVID-19 pandemic can also be observed as most crimes have seen an overall drop in 2020. For example, theft without violence saw a sharp drop and its levels have not gone back to 2019 levels of theft.

The Types of crime committed also depend on the location.

In these graphs, we can see the difference in the types of crime committed in the Paris department and in the Cher department. Cher is sparsely populated department. Here we observe that in Paris, theft without violence is the number one committed crime. This is surely due to Paris being a big city with many tourists and as such many pickpockets. This may not be the case in the Cher region with only 300’000 inhabitants. In the Cher, theft is less prevalent with the biggest Crime rate being deliberate destruction and damage.

3.1.2 Unemployment

3.1.2.1 Evolution of the unemployment rate

Let’s now take a look at our first variable: Unemployment. Unemployment in France has evolved through time and has usually been higher than unemployment in Switzerland.

We can observe a huge increase in the unemployment rate in the 90’s in France. As we know after the Second World War, European countries had an important period of economic growth, this period is known as the « 30 glorieuses ». This period matched with an important demand of workers, from Europe and from other continents. There was a need to rebuild the country and an important diversification of the jobs with help of technological progress. In the early 80’s, an important recession started in Europe, with a decrease of job creation in most of the important European countries, especially Germany, Great Britain and France. The the unemployment rate was approximately increasing by 1% a year in France in 1990, to reach the record rate of 11% in 1997. This is mainly explained by the two oil shocks in the 70’s and 80’s. The French president at the time, Francois Mitterand said in 1993 : « In terms of unemployment, we’ve tried everything and we’ve failed everything ». Following this recession France had an important decrease of unemployment rate, the explanations are multiple but the most plausible is the creation of the Euro zone and currency. This tendency lasted until the major financial crisis of 2008. In the last decade the unemployment rate has been mastered once again until the COVID 19 pandemic in 2020 which created a massive suppression of jobs in various areas such as the show industry or the food service sector.

3.1.2.2 Unemployment by department

We can observe with this graph that there’s an important disparity in unemployment rates in France. It’s also relevant that one of the departments with the higher crime rate which we’ve seen before, is also one of the departments with the higher crime rates (93). This may be an indication of what we could possibly find with the statistical regressions.

3.1.3 Immigration

We can use a Treemap to visualize the total number of immigrants by Region. Regions in France contain several Departments. ::: {.cell} ::: {.cell-output-display} ::: :::

As we can see, most immigrants are in the Paris area. These visualizations really put into perspective how big the Paris region is compared to the rest of the country.

3.2 Mapping

This map compute the crime rate in every French department and shows us the important disparity in the amount of crimes committed across the country. As we can see, Paris has a very high crime rate compared to the rest of the departments.

This map computes the unemployment percentage in every department, Unemployement is higher in the north and south as well as around Paris.

This map shows us the score in the second round of the French presidential election of 2017, which we choose to compute as a possible explanation variable. As we know Emmanuel Macron has been elected president with 66.10% so the departments overall voted for him, however it is interesting to see the departments where Marine LePen had an important share of the votes. Here those departments are represented with a darker color.

This map shows the school pass rate for each department at the end of middle school. Unfortunately, no clear trend seems to emerge from the map. Futhermore, the difference between the highest value (0.9305) and the lowest value (0.8274) is only of 10%. This means that all other values are contained within only 10%. Showing them on a map with colors tends to over-exagerate the differences between values. As such, this map has poor vizualization power.

Lastly, this map shows us the immigration rate, by department. As we can see, Immigrants are mostly located in the vicinity of the Paris region. Other notable areas of higher immigration are border regions in the south and in the east.

Here the density can be observed, with Paris and it’s region being an outlier, it is hard to see the difference between other regions. This maps shows clearly that most people in France live around Paris.

3.2.0.1 Interactive maps

We create a map using leaflet. To display another variable, simply click on the desired variable. It should be noted that, for clarity reasons, outlier values for Paris regarding Crime rates and Density of population. Have been removed from the color scale. If this hasn’t been done, it would be very difficult to see the differences in the remaining lower values. To see how not removing those values may look in a map format, please refer to the previous non-interactive maps.

In the above map, one can see the crime rate for many towns in France. To display them, simply click on the desired department. It should be noted that not all towns are represented here. Furthermore, the circle center may be slightly off from the city center, this is because the coordinates used are the center of city limits, which often does not line up with the “economical” city center for example.

3.3 Data distribution by town

We will now take a look at our variables to see how our data is distributed and the extreme values that can be found. This analysis will be conducted on our dataset with many missing(NA) values, also in part to show why we have chosen to reduce our dataset to bigger towns.

3.3.1 Total crime rate

Of our 36732 rows, 26,493 rows are equal to 0.

It appears in most of our dataset that the Paris department was an outlier and has very unique values. For example for the crime rate Paris has 112 crimes per thousand people which is almost double compared to the second highest values.

It also appears that the crime rate is very high in highly touristic places. Some places such as the winter sports resort town of Morzine or Saint-Tropez in the south have a low population during the year but see a large influx of tourists during the summer or winter. The criminality there is high because the total number of crimes is high compared to a low overall population. Same observation for the communes where French airports are located such as Roissy-Charles-de-Gaulle Airport and the Lyon airport. All crimes reported in the airport are counted as committed inside the town, with millions of passengers each year, the crime rate is inflated. The town of Chessy is also for example home to Disneyland Paris, and sees a high crime rate, with the same reasons as before.

3.3.2 Total population

The population dataset also offers some surprises when we found that there were some villages with a population equal to zero. We realized after that that it concerns the « Villages morts pour la France », some villages have been totally destroyed during the First World War and the French government decided to keep the memory by naming a mayor for those villages, responsible for the well being of the sanctuary.

3.3.3 Density

The density variable also has her own unique values. We found that some large towns in terms of population may have a low density due to the presence of natural reserves and forests on their territory. For example, the city of Arles, with a population of 50,454 and a density of only 69.388 habitants per square km. This is due to the presence of the Camargue natural reserve. The most dense towns are usually in the Paris region, with a smaller area. This is the case for Levallois-Perret, bordering Paris and our most dense town.

3.3.4 Score of Lepen

The scores of Lepen are almost normally distributed. Most of her high scores are usually in the northern regions.

3.3.5 Poverty

Towns with higher levels of poverty are located mostly in the northern parts of France or in the Paris region. This is not surprising.

3.3.6 Intensity Poverty

This variable is somewhat more surprising, while low intensity of poverty does not seem to be located anywhere in particular, Areas with higher poverty are almost exclusively located in the Ain Department, near the Swiss Border in Geneva. Among the ten towns with a high intensity of poverty, only one is not located near Switzerland, it is instead located on the border with Luxemburg.

The explanation for this may may be that many inhabitants from border regions work in the neighbouring country, and said regions often have a low unemployment rate. In the case of Switzerland and Luxemburg, these country offer higher salaries than Switzerland. As such, people below the poverty line are further away from the poverty line, resulting in a high intensity of poverty.

3.3.7 No diploma

Towns with a high level people without a diploma are mostly located in former industrial towns and rural areas, where not having a diploma was never a problem. As work usually started at a younger age and was passed down, obtaining a diploma was usally not needed. For example, the city of Behren-les-forbach, with about 30% of people over 25 without a diploma is a former mining town. Here work was provided by the mining company, without a need for any diploma. As those mines closed all over France, workers were often left without any suitable alternative to work. This explains in part the current poverty level of 51%.

3.3.8 Immigration rate

Surprisingly here, the town with the most immigrants is Taizé, a religious community where many Christian people from all over the world live. This reminds us that immigrants may come into a foreign country for many reasons and image that we may have of immigration doesn’t always apply. Other towns with a high immigration rate are either located in the Paris region or bordering another country.

3.4 Clustering

We will now run a cluster analysis of each department, we want to see if some departments have simmilarities.

First, we compute the correlation matrix.

	Unemp_2019	Crime_rate_1k	Pass_rate	Lepen_score	Immig_rate	Density_2019
Unemp_2019	1.0000	0.2554	-0.4734	0.5347	0.1254	-0.0839
Crime_rate_1k	0.2554	1.0000	-0.2235	-0.2515	0.6863	0.7659
Pass_rate	-0.4734	-0.2235	1.0000	-0.3176	-0.2521	0.0489
Lepen_score	0.5347	-0.2515	-0.3176	1.0000	-0.2983	-0.4813
Immig_rate	0.1254	0.6863	-0.2521	-0.2983	1.0000	0.5753
Density_2019	-0.0839	0.7659	0.0489	-0.4813	0.5753	1.0000

Density is highly correlated with Crime Rates. In contrast, Density and unemployment have a very low correlation. Other interesting Correlations are the one with Immigration and Crime rates, Unemployment and the Score of Marine Lepen.

3.4.1 Hierarchical clustering

We now compute the distance matrix. Using this distance matrix, we run a hierarchical clustering analysis using the average distance method. We get the following dendrogram:

This graph tells us that observation 76 stands out from the rest, Observation 76 is actually Paris. Paris is a clear outlier. Based on this graph we could try to use 3 clusters.

We can now visualize this using our 2 most correlated variables, Crime rates and Density.

This is quite interesting, our 3 clusters are Paris, the Paris region and the rest of France. We can also see this in a map format:

3.4.2 K-means clustering

To further our analysis and chose a number of clusters, we will run K-means analysis with n-start = 25. Our data will also be scaled to reduce differences. We will run the same clustering 25 times with different centers as we initialize. Running it 25 times will ensure that we get the best clusters we can.

We get the following scree plot from our analysis:

Our optimal number of clusters seems to be 6 clusters. We’ll visualize this using a cluster plot and a facet wrap of Le Pen scores and Unemployment rate.

As before, we will map it for more clarity.

Overall, during this clustering exercise, our main learning is that Paris and the agglomeration around are very big compared to the rest of France, so much that it tends skew the analysis.

4 Regression

4.1 Correlations

To see how our data may be correlated, we will start by running a correlation analysis.

	Total_pop	Total_crime	Rate_per_1k	Burglary	Assault	Damage	Drugs	Theft	Density_2019	Lepen	Win_Lepen	Povrety_2019	Intensity_povrety	No_diploma_rate1k	Immig_rate	Unemp_2019
Total_pop	1.0000	0.9671	0.0746	0.0762	0.1671	0.1658	0.1280	0.1044	0.3751	-0.1401	-0.0469	0.1412	0.1992	0.0168	0.2095	0.0000
Total_crime	0.9671	1.0000	0.0858	0.0514	0.1185	0.1208	0.0886	0.1092	0.2877	-0.0904	-0.0270	0.0914	0.1291	0.0085	0.1344	0.0039
Rate_per_1k	0.0746	0.0858	1.0000	0.2993	0.5862	0.6278	0.3172	0.8538	0.0827	0.0550	0.0477	0.1556	0.2207	0.1035	0.1838	0.1307
Burglary	0.0762	0.0514	0.2993	1.0000	0.1047	0.1234	0.0361	0.0260	0.1682	-0.0796	-0.0183	-0.0556	0.1855	-0.1342	0.2590	0.1296
Assault	0.1671	0.1185	0.5862	0.1047	1.0000	0.7038	0.4590	0.5342	0.2457	0.1082	0.1123	0.5666	0.5355	0.4131	0.3636	0.1737
Damage	0.1658	0.1208	0.6278	0.1234	0.7038	1.0000	0.4272	0.4672	0.2213	0.0609	0.0755	0.5140	0.5120	0.3452	0.3094	0.1791
Drugs	0.1280	0.0886	0.3172	0.0361	0.4590	0.4272	1.0000	0.2415	0.2018	-0.0638	-0.0207	0.3143	0.3060	0.2028	0.3055	0.0225
Theft	0.1044	0.1092	0.8538	0.0260	0.5342	0.4672	0.2415	1.0000	0.0935	-0.0299	-0.0166	0.0720	0.0975	0.0373	0.1146	0.0265
Density_2019	0.3751	0.2877	0.0827	0.1682	0.2457	0.2213	0.2018	0.0935	1.0000	-0.2946	-0.0786	0.2033	0.3044	0.0420	0.4991	-0.0055
Lepen	-0.1401	-0.0904	0.0550	-0.0796	0.1082	0.0609	-0.0638	-0.0299	-0.2946	1.0000	0.6723	0.3046	0.0611	0.4016	-0.1805	0.5327
Win_Lepen	-0.0469	-0.0270	0.0477	-0.0183	0.1123	0.0755	-0.0207	-0.0166	-0.0786	0.6723	1.0000	0.2572	0.1031	0.2200	-0.1019	0.4410
Povrety_2019	0.1412	0.0914	0.1556	-0.0556	0.5666	0.5140	0.3143	0.0720	0.2033	0.3046	0.2572	1.0000	0.6795	0.7382	0.4191	0.3587
Intensity_povrety	0.1992	0.1291	0.2207	0.1855	0.5355	0.5120	0.3060	0.0975	0.3044	0.0611	0.1031	0.6795	1.0000	0.4265	0.5341	0.2415
No_diploma_rate1k	0.0168	0.0085	0.1035	-0.1342	0.4131	0.3452	0.2028	0.0373	0.0420	0.4016	0.2200	0.7382	0.4265	1.0000	0.3073	0.2141
Immig_rate	0.2095	0.1344	0.1838	0.2590	0.3636	0.3094	0.3055	0.1146	0.4991	-0.1805	-0.1019	0.4191	0.5341	0.3073	1.0000	-0.0214
Unemp_2019	0.0000	0.0039	0.1307	0.1296	0.1737	0.1791	0.0225	0.0265	-0.0055	0.5327	0.4410	0.3587	0.2415	0.2141	-0.0214	1.0000

We first decide to compute the correlations with all of our variables dividing « crimes » in every crime categories to see if there are interesting correlation links to explain them. As we can see Burglaries are highly correlated with Immigration rate and Unemployment rate and negatively correlated to the No diploma rate. On the other side Assault and Battery is mainly correlated to poverty rate and No Diploma rate. Intentional Damage behave the same as Assault and Battery, it is also interesting to observe that voting for Marin Lepen is positively correlated with these two variables. For Drug use we observe that the two more correlated variables are Immigration and Unemployment rate. And finally we observe that Theft without violence only has small correlations with all of our explicative variables.

4.2 Linear regression

We started by scaling all variables, expressing them in percentages except for density (people/km^2) and Win_Lepen (a binary variable ranging from 0 to 1). The initial regression included the average of all crimes and potential explanatory variables. Subsequently, we conducted a stepwise regression AIC test to identify indispensable variables, leading to the removal of Win_Lepen, Intensity Poverty, Poverty, and No Diploma Rate.

4.2.1 All crimes

The initial regression includes the average of all crimes and potential explanatory variables. With a formula as seen below:

\(RatePer1k = \beta_0 + \beta_1 Density + \beta_2WinLepen + \beta_3Poverty + \beta_4NoDiplomaRate +\) \(\beta_5ImmigrationRate + \beta_5UnemploymentRate + \beta_6IntensityPoverty + \beta_6 Total Pop\)

Subsequently, we conducted a stepwise regression AIC test to identify indispensable variables, leading to the removal of Win_Lepen, Intensity Poverty, Poverty, and No Diploma Rate.

Code

step(Assault_model_full, scope = list(lower= Assault_model_null, upper=Assault_model_full),
     direction = "backward")
## Start:  AIC=7498
## Assault ~ Density_2019 + Povrety_2019 + No_diploma_rate1k + Immig_rate + 
##     Unemp_2019 + Win_Lepen + Intensity_povrety
## 
##                     Df Sum of Sq   RSS  AIC
## - Win_Lepen          1         2 22479 7496
## <none>                           22478 7498
## - Unemp_2019         1        10 22487 7498
## - Immig_rate         1        24 22502 7501
## - No_diploma_rate1k  1        33 22510 7503
## - Density_2019       1       180 22657 7535
## - Intensity_povrety  1       963 23441 7703
## - Povrety_2019       1      1102 23580 7732
## 
## Step:  AIC=7496
## Assault ~ Density_2019 + Povrety_2019 + No_diploma_rate1k + Immig_rate + 
##     Unemp_2019 + Intensity_povrety
## 
##                     Df Sum of Sq   RSS  AIC
## - Unemp_2019         1         8 22487 7496
## <none>                           22479 7496
## - Immig_rate         1        23 22503 7499
## - No_diploma_rate1k  1        34 22513 7502
## - Density_2019       1       179 22658 7533
## - Intensity_povrety  1       961 23441 7701
## - Povrety_2019       1      1119 23598 7734
## 
## Step:  AIC=7496
## Assault ~ Density_2019 + Povrety_2019 + No_diploma_rate1k + Immig_rate + 
##     Intensity_povrety
## 
##                     Df Sum of Sq   RSS  AIC
## <none>                           22487 7496
## - Immig_rate         1        30 22517 7500
## - No_diploma_rate1k  1        37 22524 7502
## - Density_2019       1       179 22667 7533
## - Intensity_povrety  1       954 23441 7699
## - Povrety_2019       1      1152 23639 7740
## 
## Call:
## lm(formula = Assault ~ Density_2019 + Povrety_2019 + No_diploma_rate1k + 
##     Immig_rate + Intensity_povrety, data = Crime_type)
## 
## Coefficients:
##       (Intercept)       Density_2019       Povrety_2019  
##         -2.571359           0.000126          10.040860  
## No_diploma_rate1k         Immig_rate  Intensity_povrety  
##          3.307326           1.695249           8.234935

The goal of the stepwise regression is to only keep useful variables in our regression. Here, the AIC is an indicator that goes up if the regression has some unnecessary variables, the stepwise will check at each step and remove variables if they do not meet the criteria. When the AIC test is at its lowest it means that all the unnecessary variables have been removed.

We find this reduced model:

Characteristic	Full model			Reduced model
	Beta1	SE2	VIF2	Beta1	SE2	VIF2
Density_2019	0.00	0.000	1.5	0.00	0.000	1.5
Win_Lepen	0.06	0.122	1.3
Povrety_2019	-1.7*	0.823	3.8
No_diploma_rate1k	0.89	1.47	2.4
Immig_rate	5.9***	0.854	1.8	8.8***	0.732	1.3
Unemp_2019	18***	2.72	1.4	23***	2.32	1.0
Intensity_povrety	5.5***	0.714	2.2
Total_pop	0.00*	0.000	1.2	0.00**	0.000	1.2
1 *p<0.05; **p<0.01; ***p<0.001
2 SE = Standard Error, VIF = Variance Inflation Factor

The remaining significant variables, including Immigration, Unemployment, TotalPop, and Density, passed the Variance Inflation Factor test, ensuring no issues of multicollinearity (VIF score < 5). However, we noted a low Adjusted R-Squared, indicating that our model explained only a small proportion of the data. To address this, we performed separate regressions for the five most prevalent crime categories in France: Assault and Battery, Burglaries, Wilful Damage and Destructions, Drug Use, and Theft without Violence.

It’s crucial to emphasize that the stepwise regression with AIC mentioned earlier was applied to each of our regressions. The regressions presented below have been refined to exclude irrelevant variables based on this test.

4.2.2 Assault and Battery

We run a stepwise regression with Asssault and Battery as the dependent variable:

Characteristic	Beta1	SE2	VIF2
Density_2019	0.00***	0.000	1.4
Povrety_2019	10***	0.632	3.4
No_diploma_rate1k	3.3**	1.17	2.3
Immig_rate	1.7*	0.666	1.7
Intensity_povrety	8.2***	0.570	2.2
1 *p<0.05; **p<0.01; ***p<0.001
2 SE = Standard Error, VIF = Variance Inflation Factor

In the Assault and Battery regression, Density, Poverty, No_Diploma_Rate, and Intensity_Poverty emerged as significant variables, with Intensity_Poverty having the most substantial impact (beta of 8.235). The VIF test confirmed the absence of multicollinearity, and the Adjusted R-Squared improved to around 0.37.

Our final formula is the following: \(AssaultandBattery = \beta_0 + \beta_1 Density + \beta_2Poverty + \beta_3NoDiplomaRate +\) \(\beta_4ImmigrationRate + \beta_5IntensityPoverty\)

4.2.3 Burglaries

We run the regression with Burglaries and find this reduced model:

Characteristic	Beta1	SE2	VIF2
Povrety_2019	-20***	1.43	3.7
No_diploma_rate1k	-18***	2.52	2.3
Immig_rate	27***	1.34	1.5
Unemp_2019	71***	4.43	1.2
Intensity_povrety	16***	1.24	2.2
1 *p<0.05; **p<0.01; ***p<0.001
2 SE = Standard Error, VIF = Variance Inflation Factor

For the Burglaries regression, all variables remained highly significant. Notably, Poverty and No_Diploma_Rate negatively affected predictions (-19.624/-17.833), while Unemployment had the most significant positive impact (71.162), followed by Immigration (26.937) and Intensity_Poverty (16.391). The VIF test assured us of no multicollinearity issues, though the Adjusted R-Squared was approximately 0.19.

Our final formula is the following: \(Burglary = \beta_0 + \beta_1Poverty + \beta_2NoDiplomaRate + \beta_3ImmigrationRate +\) \(\beta_4UnemploymentRate + \beta_5IntensityPoverty\)

4.2.4 Intentional damage and destructions

We run the model with Damage and Destruction:

Characteristic	Beta1	SE2	VIF2
Povrety_2019	16***	0.806	2.0
Win_Lepen	-0.36*	0.150	1.1
Intensity_povrety	15***	0.901	2.0
Density_2019	0.00***	0.000	1.1
1 *p<0.05; **p<0.01; ***p<0.001
2 SE = Standard Error, VIF = Variance Inflation Factor

In the Intentional Damage and Destruction regression, Poverty, Intensity_Poverty, Density, and somewhat less so Win_Lepen were highly significant. Both Poverty and Intensity had a substantial positive impact (~15), while the impact of Win_Lepen was negligible. The VIF test once again confirmed no multicollinearity.

Our final formula is the following: \(DamageDestruction = \beta_0 + \beta_1Poverty + \beta_2WinLepen + \beta_3IntensityPoverty + \beta_4 Density\)

4.2.5 Drug Use

We run the model with Drug use as our dependent variable:

Characteristic	Beta1	SE2	VIF2
Density_2019	0.00***	0.000	1.4
Povrety_2019	7.8***	0.781	3.7
No_diploma_rate1k	-2.8*	1.39	2.4
Immig_rate	6.2***	0.811	1.8
Unemp_2019	-9.6***	2.59	1.4
Win_Lepen	-0.32**	0.116	1.3
Intensity_povrety	2.8***	0.678	2.2
1 *p<0.05; **p<0.01; ***p<0.001
2 SE = Standard Error, VIF = Variance Inflation Factor

For the Drug Use regression, the AIC stepwise regression test retained all variables except TotalPop. Poverty and Immigration Rate had the most significant positive impacts (7.769/6.178), while Unemployment had a considerable negative impact (-9.551). The VIF test indicated no multicollinearity issues.

Our final formula is the following: \(DrugUse = \beta_0 + \beta_1 Density + \beta_2Poverty + \beta_3NoDiplomaRate + \beta_4ImmigrationRate +\) \(\beta_5UnemploymentRate + \beta_6WinLepen + \beta_7IntensityPoverty\)

4.2.6 Theft without violence

We run the model with Theft as our independent variable:

Characteristic	Beta1	SE2	VIF2
Immig_rate	21***	5.79	1.8
Win_Lepen	-1.2	0.828	1.3
Density_2019	0.00**	0.000	1.3
Unemp_2019	31	18.4	1.3
Intensity_povrety	10*	4.12	1.5
1 *p<0.05; **p<0.01; ***p<0.001
2 SE = Standard Error, VIF = Variance Inflation Factor

In the Theft without Violence regression, Immigration Rate and Density were the only highly significant variables. The positive impact of Immigration_Rate was substantial (~20), but the low Adjusted R-Squared suggests challenges in interpreting results. Nevertheless, the VIF test confirmed no multicollinearity issues in this regression.”

Our final formula is the following: \(Theft = \beta_0 + \beta_1 Density + \beta_2WinLepen + \beta_3ImmigrationRate +\) \(\beta_4UnemploymentRate + \beta_5IntensityPoverty\)

4.3 Discussing our results

4.3.0.1 Unemployment

In our initial research, we wanted to understand if Unemployment could predict Crime in France. When we analyzed all types of crimes together, we found a connection between Unemployment and Crime. However, our findings lacked precision due to a low Adjusted R-Squared value. Upon further investigation, looking at each crime type separately, we found some interesting results. Unemployment’s influence on predicting Crime appeared to change depending on the specific crime category. For instance, in cases like Assault and Battery and Intentional Damage and Destruction, Unemployment didn’t seem to significantly impact the calculations. This was counterintuitive for our expectations. Moreover, concerning Burglaries and Theft without Violence, Unemployment emerged as a more critical factor. It became the primary predictor for the occurrence of these crime types.

4.3.0.2 Poverty

We then delved into the potential impact of Poverty on predicting Crime in France. We examined two datasets: one reflecting the number of people affected by poverty in each town (‘Poverty_2019’) and another showing how far individuals were to the poverty threshold (‘Intensity_Poverty’). Again, we analyzed each crime category separately for accuracy. In certain crimes, such as Assault and Battery and Intentional Damage and Destruction, poverty and its intensity significantly influenced crime rates. We then faced a very surprising result when computing the regression with the Burglaries, in this case, Poverty seems to have an important negative impact while the Intensity of Poverty has a postive impact. The VIF factor was also under 5 meaning that multicollinearity was not the reason for that result. This leads us to the interpretation that when talking about Burglaries, the amount of poor people in a town would have a negative impact since Richer neighbourhoods are more likely to be burglarized. In the case of Intensity of Poverty, this varibale is not an indicator of how many people are poor, but rather how poor they are we could assume that the poorest people who burglarize. Regarding Drug Use crimes, Poverty emerged as a significant predictor, but with notably lesser impact compared to previous regressions. Similarly, when examining Theft without Violence, while ‘Intensity_Poverty’ showed promise, it didn’t reach statistical significance. In essence, while Poverty and its intensity play substantial roles in predicting criminal behavior in France, their impact varies based on the specific crime type, often leading to conflicting trends.

4.3.0.3 Immigration

Moving on to the impact of immigration on crime rates in France, our analyses revealed nuanced outcomes. For instance, in ‘Assault and Battery’ and ‘Intentional Damage and Destruction’ cases, the Immigration variable appeared insignificant, with minimal influence. However, in Burglaries and Drug-related crimes, Immigration seemed to exert a significant effect. Nonetheless, its impact on ‘Theft without Violence’ remained uncertain due to a low Adjusted R-Squared value, cautioning against definitive conclusions. Overall, while Immigration played a significant role in some models, its impact varied considerably among different crime types.

4.3.0.4 Population Density

We also explored the role of Population Density in predicting criminal activity. Interestingly, Density didn’t seem to affect Burglaries much, possibly because criminals find it easier to operate in less crowded, rural areas, For example, in the department of Gironde, where Bordeaux is located, the city with the most burglaries. Concerning the remaining crime types, Density is always significant, but has a small coefficient. We observe that some of the most criminal cities in France, such as Marseille or Lyon are not very dense. This explains why Density may not be the best factor to explain Crime.

4.3.0.5 Education level

In terms of Education level, the data suggests that the population without a diploma doesn’t always have the highest crime rates, which might be unexpected. For instance, in the case of Assault and Battery, our findings indicate that crime rates do relate to education level, but the impact is minimal. In many areas across France, especially in the northeastern mining regions, there’s a large number of individuals without diplomas, yet the crime rates there is remarkably low. However, in La Courneuve (93) a city in the Paris region with a notably high crime rate, there is a considerable proportion of people without a diploma. Furthermore, the absence of a diploma seems to have a negative correlation with Burglary rates, with a significant factor of -17. This isn’t surprising as it is a wealthier region, implying that cities with a higher education level might be the primary targets for burglaries. Regarding Drug Use, it was unexpected to see that the absence of a diploma negatively impacts the regression. This could be explained by acknowledging that drug use is widespread among the population in France, affecting even the more educated regions.

4.3.0.6 Election results

Lastly, we explored whether there was a connection between voting patterns in the 2017 presidential election and the crime rates. We aimed to understand if the way people voted affected the crimes committed in those areas and if political behavior correlated with observed criminality. While the AIC eliminated our Voting variables in most of our regressions, an interesting negative relationship emerged between Marine Le Pen “winning” the city and both Drug Use and Intentional Damage and Destruction regressions. For instance, if a city voted in majority for Marine Le Pen in 2017, it had on average an Intentional Damage and Destruction rate 0.362 lower than a city that voted against her. That decrease was 0.319. in the Drug Use category.. However, apart from these instances, the Le Pen variables weren’t significant enough to draw conclusive remarks for our other variables.

4.3.0.7 Correlations

Regarding the correlation between our variables, we observe an absence of multicollinearity which is very positive. In fact the Variance Inflation Factor shows us that there is no consequent amount of information contained by one of the variables that is also contained by another, the opposite would have been a problem for the quality of our regressions. Nevertheless with the correlation map we discovered that there was an important positive link between Unemployment and Le Pen vote, as much as an important one between Immigration and Density, Poverty and No Diploma rate. Overall, there are some obvious links between our variables but they seem to behave slightly differently in the regressions.

5 Conclusion and limits

This project aimed to predict criminality in various aspects across France using several explanatory variables. However, significant limitations exist.

Firstly, France exhibits substantial disparities between its departments and cities concerning crime rates, education levels, unemployment, and immigration. Interpreting these results at a national level requires caution. A more granular analysis at the department level, though reducing data volume and calculation precision, could provide more accurate insights, considering the unique characteristics of each territory.

Another limitation is the lack of detailed information in our crime dataset, particularly the origin city of the perpetrator. Not all crimes are committed by residents of the city where the crime occurs, and this nuance is not accounted for in our analysis. To gain a more comprehensive understanding of trends, having the same crime dataset for multiple years could be beneficial. The current study only provides a snapshot of 2019, while crime dynamics evolve through the years, as illustrated in previous graphics. Indeed, incorporating data from multiple years, especially around election years, could offer valuable insights into the impact of political changes on criminality. Analyzing trends over time, particularly in relation to the LePen variable, would provide a more nuanced understanding of how government shifts influence crime rates. Moreover, municipal elections may better capture local sentiments and dynamics, potentially offering a more accurate reflection of the community’s stance, which could have a distinct influence on crime behavior. Including such localized political data, instead of national votation data would enhance the precision and relevance of the analysis.

Nevertheless, this project provides an overview of the distribution and characteristics of crime across the French territory, its evolution, and its main drivers. It allowed us to explore the distribution of key socio-economic indicators at both the departmental and local levels. We hope that this work will be as useful to you as it has been enjoyable for us to prepare.